arXiv Daily Digest - 2026-02-17

Articulated object manipulation is essential for various real-world robotic tasks, yet generalizing across diverse objects remains a major challenge. A key to generalization lies in understanding functional parts (e.g., door handles and knobs), which indicate where and how to manipulate across diverse object categories and shapes. Previous works attempted to achieve generalization by introducing foundation features, while these features are mostly 2D-based and do not specifically consider functional parts. When lifting these 2D features to geometry-profound 3D space, challenges arise, such as long runtimes, multi-view inconsistencies, and low spatial resolution with insufficient geometric information. To address these issues, we propose Part-Aware 3D Feature Field (PA3FF), a novel dense 3D feature with part awareness for generalizable articulated object manipulation. PA3FF is trained by 3D part proposals from a large-scale labeled dataset, via a contrastive learning formulation. Given point clouds as input, PA3FF predicts a continuous 3D feature field in a feedforward manner, where the distance between point features reflects the proximity of functional parts: points with similar features are more likely to belong to the same part. Building on this feature, we introduce the Part-Aware Diffusion Policy (PADP), an imitation learning framework aimed at enhancing sample efficiency and generalization for robotic manipulation. We evaluate PADP on several simulated and real-world tasks, demonstrating that PA3FF consistently outperforms a range of 2D and 3D representations in manipulation scenarios, including CLIP, DINOv2, and Grounded-SAM. Beyond imitation learning, PA3FF enables diverse downstream methods, including correspondence learning and segmentation tasks, making it a versatile foundation for robotic manipulation. Project page: https://pa3ff.github.io

Large language models are increasingly used to answer and verify scientific claims, yet existing evaluations typically assume that a model must always produce a definitive answer. In scientific settings, however, unsupported or uncertain conclusions can be more harmful than abstaining. We study this problem through an abstention-aware verification framework that decomposes scientific claims into minimal conditions, audits each condition against available evidence using natural language inference (NLI), and selectively decides whether to support, refute, or abstain. We evaluate this framework across two complementary scientific benchmarks: SciFact and PubMedQA, covering both closed-book and open-domain evidence settings. Experiments are conducted with six diverse language models, including encoder-decoder, open-weight chat models, and proprietary APIs. Across all benchmarks and models, we observe that raw accuracy varies only modestly across architectures, while abstention plays a critical role in controlling error. In particular, confidence-based abstention substantially reduces risk at moderate coverage levels, even when absolute accuracy improvements are limited. Our results suggest that in scientific reasoning tasks, the primary challenge is not selecting a single best model, but rather determining when available evidence is sufficient to justify an answer. This work highlights abstention-aware evaluation as a practical and model-agnostic lens for assessing scientific reliability, and provides a unified experimental basis for future work on selective reasoning in scientific domains. Code is available at https://github.com/sabdaljalil2000/ai4science .

With the significant expansion of the context window in Large Language Models (LLMs), these models are theoretically capable of processing millions of tokens in a single pass. However, research indicates a significant gap between this theoretical capacity and the practical ability of models to robustly utilize information within long contexts, especially in tasks that require a comprehensive understanding of numerous details. This paper evaluates the performance of four state-of-the-art models (Grok-4, GPT-4, Gemini 2.5, and GPT-5) on long short-context tasks. For this purpose, three datasets were used: two supplementary datasets for retrieving culinary recipes and math problems, and a primary dataset of 20K social media posts for depression detection. The results show that as the input volume on the social media dataset exceeds 5K posts (70K tokens), the performance of all models degrades significantly, with accuracy dropping to around 50-53% for 20K posts. Notably, in the GPT-5 model, despite the sharp decline in accuracy, its precision remained high at approximately 95%, a feature that could be highly effective for sensitive applications like depression detection. This research also indicates that the "lost in the middle" problem has been largely resolved in newer models. This study emphasizes the gap between the theoretical capacity and the actual performance of models on complex, high-volume data tasks and highlights the importance of metrics beyond simple accuracy for practical applications.

Unified Multimodal Large Language Models (MLLMs) require a visual representation that simultaneously supports high-fidelity reconstruction, complex semantic extraction, and generative suitability. However, existing visual tokenizers typically struggle to satisfy these conflicting objectives within a single framework. In this paper, we introduce UniWeTok, a unified discrete tokenizer designed to bridge this gap using a massive binary codebook ($\mathit{2^{128}}$). For training framework, we introduce Pre-Post Distillation and a Generative-Aware Prior to enhance the semantic extraction and generative prior of the discrete tokens. In terms of model architecture, we propose a convolution-attention hybrid architecture with the SigLu activation function. SigLu activation not only bounds the encoder output and stabilizes the semantic distillation process but also effectively addresses the optimization conflict between token entropy loss and commitment loss. We further propose a three-stage training framework designed to enhance UniWeTok's adaptability cross various image resolutions and perception-sensitive scenarios, such as those involving human faces and textual content. On ImageNet, UniWeTok achieves state-of-the-art image generation performance (FID: UniWeTok 1.38 vs. REPA 1.42) while requiring a remarkably low training compute (Training Tokens: UniWeTok 33B vs. REPA 262B). On general-domain, UniWeTok demonstrates highly competitive capabilities across a broad range of tasks, including multimodal understanding, image generation (DPG Score: UniWeTok 86.63 vs. FLUX.1 [Dev] 83.84), and editing (GEdit Overall Score: UniWeTok 5.09 vs. OmniGen 5.06). We release code and models to facilitate community exploration of unified tokenizer and MLLM.

Spatial transcriptomics (ST) provides spatially resolved measurements of gene expression, enabling characterization of the molecular landscape of human tissue beyond histological assessment as well as localized readouts that can be aligned with morphology. Concurrently, the success of multimodal foundation models that integrate vision with complementary modalities suggests that morphomolecular coupling between local expression and morphology can be systematically used to improve histological representations themselves. We introduce Spatial Expression-Aligned Learning (SEAL), a vision-omics self-supervised learning framework that infuses localized molecular information into pathology vision encoders. Rather than training new encoders from scratch, SEAL is designed as a parameter-efficient vision-omics finetuning method that can be flexibly applied to widely used pathology foundation models. We instantiate SEAL by training on over 700,000 paired gene expression spot-tissue region examples spanning tumor and normal samples from 14 organs. Tested across 38 slide-level and 15 patch-level downstream tasks, SEAL provides a drop-in replacement for pathology foundation models that consistently improves performance over widely used vision-only and ST prediction baselines on slide-level molecular status, pathway activity, and treatment response prediction, as well as patch-level gene expression prediction tasks. Additionally, SEAL encoders exhibit robust domain generalization on out-of-distribution evaluations and enable new cross-modal capabilities such as gene-to-image retrieval. Our work proposes a general framework for ST-guided finetuning of pathology foundation models, showing that augmenting existing models with localized molecular supervision is an effective and practical step for improving visual representations and expanding their cross-modal utility.

Paralinguistic and non-linguistic aspects of speech strongly influence listener impressions. While most research focuses on absolute impression scoring, this study investigates relative voice impression estimation (RIE), a framework for predicting the perceptual difference between two utterances from the same speaker. The estimation target is a low-dimensional vector derived from subjective evaluations, quantifying the perceptual shift of the second utterance relative to the first along an antonymic axis (e.g., ``Dark--Bright''). To isolate expressive and prosodic variation, we used recordings of a professional speaker reading a text in various styles. We compare three modeling approaches: classical acoustic features commonly used for speech emotion recognition, self-supervised speech representations, and multimodal large language models (MLLMs). Our results demonstrate that models using self-supervised representations outperform methods with classical acoustic features, particularly in capturing complex and dynamic impressions (e.g., ``Cold--Warm'') where classical features fail. In contrast, current MLLMs prove unreliable for this fine-grained pairwise task. This study provides the first systematic investigation of RIE and demonstrates the strength of self-supervised speech models in capturing subtle perceptual variations.

Effective exploration is a key challenge in reinforcement learning for large language models: discovering high-quality trajectories within a limited sampling budget from the vast natural language sequence space. Existing methods face notable limitations: GRPO samples exclusively from the root, saturating high-probability trajectories while leaving deep, error-prone states under-explored. Tree-based methods blindly disperse budgets across trivial or unrecoverable states, causing sampling dilution that fails to uncover rare correct suffixes and destabilizes local baselines. To address this, we propose Deep Dense Exploration (DDE), a strategy that focuses exploration on $\textit{pivots}$-deep, recoverable states within unsuccessful trajectories. We instantiate DDE with DEEP-GRPO, which introduces three key innovations: (1) a lightweight data-driven utility function that automatically balances recoverability and depth bias to identify pivot states; (2) local dense resampling at each pivot to increase the probability of discovering correct subsequent trajectories; and (3) a dual-stream optimization objective that decouples global policy learning from local corrective updates. Experiments on mathematical reasoning benchmarks demonstrate that our method consistently outperforms GRPO, tree-based methods, and other strong baselines.

Existing multimodal document question answering methods universally adopt a supply-side ingestion strategy: running a Vision-Language Model (VLM) on every page during indexing to generate comprehensive descriptions, then answering questions through text retrieval. However, this "pre-ingestion" approach is costly (a 113-page engineering drawing package requires approximately 80,000 VLM tokens), end-to-end unreliable (VLM outputs may fail to be correctly retrieved due to format mismatches in the retrieval infrastructure), and irrecoverable once it fails. This paper proposes the Deferred Visual Ingestion (DVI) framework, adopting a demand-side ingestion strategy: the indexing phase performs only lightweight metadata extraction, deferring visual understanding to the moment users pose specific questions. DVI's core principle is "Index for locating, not understanding"--achieving page localization through structured metadata indexes and BM25 full-text search, then sending original images along with specific questions to a VLM for targeted analysis. Experiments on two real industrial engineering drawings (113 pages + 7 pages) demonstrate that DVI achieves comparable overall accuracy at zero ingestion VLM cost (46.7% vs. 48.9%), an effectiveness rate of 50% on visually necessary queries (vs. 0% for pre-ingestion), and 100% page localization (98% search space compression). DVI also supports interactive refinement and progressive caching, transforming the "QA accuracy" problem into a "page localization" problem--once the correct drawing page is found, obtaining the answer becomes a matter of interaction rounds.

Detecting prompt injection and jailbreak attacks is critical for deploying LLM-based agents safely. As agents increasingly process untrusted data from emails, documents, tool outputs, and external APIs, robust attack detection becomes essential. Yet current evaluation practices and production systems have fundamental limitations. We present a comprehensive analysis using a diverse benchmark of 18 datasets spanning harmful requests, jailbreaks, indirect prompt injections, and extraction attacks. We propose Leave-One-Dataset-Out (LODO) evaluation to measure true out-of-distribution generalization, revealing that the standard practice of train-test splits from the same dataset sources severely overestimates performance: aggregate metrics show an 8.4 percentage point AUC inflation, but per-dataset gaps range from 1% to 25% accuracy-exposing heterogeneous failure modes. To understand why classifiers fail to generalize, we analyze Sparse Auto-Encoder (SAE) feature coefficients across LODO folds, finding that 28% of top features are dataset-dependent shortcuts whose class signal depends on specific dataset compositions rather than semantic content. We systematically compare production guardrails (PromptGuard 2, LlamaGuard) and LLM-as-judge approaches on our benchmark, finding all three fail on indirect attacks targeting agents (7-37% detection) and that PromptGuard 2 and LlamaGuard cannot evaluate agentic tool injection due to architectural limitations. Finally, we show that LODO-stable SAE features provide more reliable explanations for classifier decisions by filtering dataset artifacts. We release our evaluation framework at https://github.com/maxf-zn/prompt-mining to establish LODO as the appropriate protocol for prompt attack detection research.

Clinical decision-making requires nuanced reasoning over heterogeneous evidence and traceable justifications. While recent LLM multi-agent systems (MAS) show promise, they largely optimise for outcome accuracy while overlooking process-grounded reasoning aligned with clinical standards. One critical real-world case of this is gene-disease validity curation, where experts must determine whether a gene is causally implicated in a disease by synthesising diverse biomedical evidence. We introduce an agent-as-tool reinforcement learning framework for this task with two objectives: (i) process-level supervision to ensure reasoning follows valid clinical pathways, and (ii) efficient coordination via a hierarchical multi-agent system. Our evaluation on the ClinGen dataset shows that with outcome-only rewards, MAS with a GRPO-trained Qwen3-4B supervisor agent substantially improves final outcome accuracy from 0.195 with a base model supervisor to 0.732, but results in poor process alignment (0.392 F1). Conversely, with process + outcome rewards, MAS with GRPO-trained supervisor achieves higher outcome accuracy (0.750) while significantly improving process fidelity to 0.520 F1. Our code is available at https://github.com/chaeeunlee-io/GeneDiseaseCurationAgents.

Sparse Mixture-of-Experts (MoE) models scale Transformers efficiently but suffer from expert overlap -- redundant representations across experts and routing ambiguity, resulting in severely underutilized model capacity. While architectural solutions like DeepSeekMoE promote specialization, they require substantial structural modifications and rely solely on intra-layer signals. In this paper, we propose two plug-and-play regularization losses that enhance MoE specialization and routing efficiency without modifying router or model architectures. First, an intra-layer specialization loss penalizes cosine similarity between experts' SwiGLU activations on identical tokens, encouraging experts to specialize in complementary knowledge. Second, a cross-layer coupling loss maximizes joint Top-$k$ routing probabilities across adjacent layers, establishing coherent expert pathways through network depth while reinforcing intra-layer expert specialization. Both losses are orthogonal to the standard load-balancing loss and compatible with both the shared-expert architecture in DeepSeekMoE and vanilla top-$k$ MoE architectures. We implement both losses as a drop-in Megatron-LM module. Extensive experiments across pre-training, fine-tuning, and zero-shot benchmarks demonstrate consistent task gains, higher expert specialization, and lower-entropy routing; together, these improvements translate into faster inference via more stable expert pathways.

Large language models (LLMs) show promise for healthcare question answering, but clinical use is limited by weak verification, insufficient evidence grounding, and unreliable confidence signalling. We propose a multi-agent medical QA framework that combines complementary LLMs with evidence retrieval, uncertainty estimation, and bias checks to improve answer reliability. Our approach has two phases. First, we fine-tune three representative LLM families (GPT, LLaMA, and DeepSeek R1) on MedQuAD-derived medical QA data (20k+ question-answer pairs across multiple NIH domains) and benchmark generation quality. DeepSeek R1 achieves the strongest scores (ROUGE-1 0.536 +- 0.04; ROUGE-2 0.226 +-0.03; BLEU 0.098 -+ 0.018) and substantially outperforms the specialised biomedical baseline BioGPT in zero-shot evaluation. Second, we implement a modular multi-agent pipeline in which a Clinical Reasoning agent (fine-tuned LLaMA) produces structured explanations, an Evidence Retrieval agent queries PubMed to ground responses in recent literature, and a Refinement agent (DeepSeek R1) improves clarity and factual consistency; an optional human validation path is triggered for high-risk or high-uncertainty cases. Safety mechanisms include Monte Carlo dropout and perplexity-based uncertainty scoring, plus lexical and sentiment-based bias detection supported by LIME/SHAP-based analyses. In evaluation, the full system achieves 87% accuracy with relevance around 0.80, and evidence augmentation reduces uncertainty (perplexity 4.13) compared to base responses, with mean end-to-end latency of 36.5 seconds under the reported configuration. Overall, the results indicate that agent specialisation and verification layers can mitigate key single-model limitations and provide a practical, extensible design for evidence-based and bias-aware medical AI.

Text-driven image and video editing can be naturally cast as inpainting problems, where masked regions are reconstructed to remain consistent with both the observed content and the editing prompt. Recent advances in test-time guidance for diffusion and flow models provide a principled framework for this task; however, existing methods rely on costly vector--Jacobian product (VJP) computations to approximate the intractable guidance term, limiting their practical applicability. Building upon the recent work of Moufad et al. (2025), we provide theoretical insights into their VJP-free approximation and substantially extend their empirical evaluation to large-scale image and video editing benchmarks. Our results demonstrate that test-time guidance alone can achieve performance comparable to, and in some cases surpass, training-based methods.

Differentiating through the solution of a quadratic program (QP) is a central problem in differentiable optimization. Most existing approaches differentiate through the Karush--Kuhn--Tucker (KKT) system, but their computational cost and numerical robustness can degrade at scale. To address these limitations, we propose dXPP, a penalty-based differentiation framework that decouples QP solving from differentiation. In the solving step (forward pass), dXPP is solver-agnostic and can leverage any black-box QP solver. In the differentiation step (backward pass), we map the solution to a smooth approximate penalty problem and implicitly differentiate through it, requiring only the solution of a much smaller linear system in the primal variables. This approach bypasses the difficulties inherent in explicit KKT differentiation and significantly improves computational efficiency and robustness. We evaluate dXPP on various tasks, including randomly generated QPs, large-scale sparse projection problems, and a real-world multi-period portfolio optimization task. Empirical results demonstrate that dXPP is competitive with KKT-based differentiation methods and achieves substantial speedups on large-scale problems.

Activation steering provides parameter-efficient control over large language models (LLMs) at inference time, but many methods rely on off-distribution supervision and discrete masking, leading to brittle interventions. We propose ROAST (Rollout-based On-distribution Activation Steering Technique), which estimates steering directions from the model's own on-distribution rollouts via ROC and avoids hard sparsification via Continuous Soft Scaling (CSS) and Grouped Mean Normalization. Our empirical analysis reveals that while activation magnitude correlates moderately with directional consistency, the variance in magnitude is significant and often disproportionate to semantic quality. This suggests that high-magnitude activations risk dominating the global steering direction if not properly normalized. To address this, ROAST employs grouped normalization to balance contributions across samples, ensuring a more robust estimation of the consensus steering direction. Across models (0.6B to 32B), ROAST consistently improves performance on diverse tasks (e.g., +9.7% on GSM8K for Qwen3-0.6B and +12.1% on TruthfulQA for GLM4-32B), and analyses show that CSS better preserves activation energy.

Traditional mechanized chestnut harvesting is too costly for small producers, non-selective, and prone to damaging nuts. Accurate, reliable detection of chestnuts on the orchard floor is crucial for developing low-cost, vision-guided automated harvesting technology. However, developing a reliable chestnut detection system faces challenges in complex environments with shading, varying natural light conditions, and interference from weeds, fallen leaves, stones, and other foreign on-ground objects, which have remained unaddressed. This study collected 319 images of chestnuts on the orchard floor, containing 6524 annotated chestnuts. A comprehensive set of 29 state-of-the-art real-time object detectors, including 14 in the YOLO (v11-13) and 15 in the RT-DETR (v1-v4) families at varied model scales, was systematically evaluated through replicated modeling experiments for chestnut detection. Experimental results show that the YOLOv12m model achieves the best mAP@0.5 of 95.1% among all the evaluated models, while the RT-DETRv2-R101 was the most accurate variant among RT-DETR models, with mAP@0.5 of 91.1%. In terms of mAP@[0.5:0.95], the YOLOv11x model achieved the best accuracy of 80.1%. All models demonstrate significant potential for real-time chestnut detection, and YOLO models outperformed RT-DETR models in terms of both detection accuracy and inference, making them better suited for on-board deployment. Both the dataset and software programs in this study have been made publicly available at https://github.com/AgFood-Sensing-and-Intelligence-Lab/ChestnutDetection.

Rapidly evolving AI exhibits increasingly strong autonomy and goal-directed capabilities, accompanied by derivative systemic risks that are more unpredictable, difficult to control, and potentially irreversible. However, current AI safety evaluation systems suffer from critical limitations such as restricted risk dimensions and failed frontier risk detection. The lagging safety benchmarks and alignment technologies can hardly address the complex challenges posed by cutting-edge AI models. To bridge this gap, we propose the "ForesightSafety Bench" AI Safety Evaluation Framework, beginning with 7 major Fundamental Safety pillars and progressively extends to advanced Embodied AI Safety, AI4Science Safety, Social and Environmental AI risks, Catastrophic and Existential Risks, as well as 8 critical industrial safety domains, forming a total of 94 refined risk dimensions. To date, the benchmark has accumulated tens of thousands of structured risk data points and assessment results, establishing a widely encompassing, hierarchically clear, and dynamically evolving AI safety evaluation framework. Based on this benchmark, we conduct systematic evaluation and in-depth analysis of over twenty mainstream advanced large models, identifying key risk patterns and their capability boundaries. The safety capability evaluation results reveals the widespread safety vulnerabilities of frontier AI across multiple pillars, particularly focusing on Risky Agentic Autonomy, AI4Science Safety, Embodied AI Safety, Social AI Safety and Catastrophic and Existential Risks. Our benchmark is released at https://github.com/Beijing-AISI/ForesightSafety-Bench. The project website is available at https://foresightsafety-bench.beijing-aisi.ac.cn/.

Multimodal Large Language Models (MLLMs) have demonstrated exceptional capabilities in high-level visual understanding. However, extending these models to fine-grained dense prediction tasks, such as semantic segmentation and depth estimation, typically necessitates the incorporation of complex, task-specific decoders and other customizations. This architectural fragmentation increases model complexity and deviates from the generalist design of MLLMs, ultimately limiting their practicality. In this work, we challenge this paradigm by accommodating standard MLLMs to perform dense predictions without requiring additional task-specific decoders. The proposed model is called DenseMLLM, grounded in the standard architecture with a novel vision token supervision strategy for multiple labels and tasks. Despite its minimalist design, our model achieves highly competitive performance across a wide range of dense prediction and vision-language benchmarks, demonstrating that a standard, general-purpose MLLM can effectively support dense perception without architectural specialization.

Large language models (LLMs) have achieved remarkable success in generating fluent and contextually appropriate text; however, their capacity to produce genuinely creative outputs remains limited. This paper posits that this limitation arises from a structural property of contemporary LLMs: when provided with rich context, the space of future generations becomes strongly constrained, and the generation process is effectively governed by near-deterministic dynamics. Recent approaches such as test-time scaling and context adaptation improve performance but do not fundamentally alter this constraint. To address this issue, we propose Algebraic Quantum Intelligence (AQI) as a computational framework that enables systematic expansion of semantic space. AQI is formulated as a noncommutative algebraic structure inspired by quantum theory, allowing properties such as order dependence, interference, and uncertainty to be implemented in a controlled and designable manner. Semantic states are represented as vectors in a Hilbert space, and their evolution is governed by C-values computed from noncommutative operators, thereby ensuring the coexistence and expansion of multiple future semantic possibilities. In this study, we implement AQI by extending a transformer-based LLM with more than 600 specialized operators. We evaluate the resulting system on creative reasoning benchmarks spanning ten domains under an LLM-as-a-judge protocol. The results show that AQI consistently outperforms strong baseline models, yielding statistically significant improvements and reduced cross-domain variance. These findings demonstrate that noncommutative algebraic dynamics can serve as a practical and reproducible foundation for machine creativity. Notably, this architecture has already been deployed in real-world enterprise environments.

Open Radio Access Networks (O-RAN) promise flexible 6G network access through disaggregated, software-driven components and open interfaces, but this programmability also increases operational complexity. Multiple control loops coexist across the service management layer and RAN Intelligent Controller (RIC), while independently developed control applications can interact in unintended ways. In parallel, recent advances in generative Artificial Intelligence (AI) are enabling a shift from isolated AI models toward agentic AI systems that can interpret goals, coordinate multiple models and control functions, and adapt their behavior over time. This article proposes a multi-scale agentic AI framework for O-RAN that organizes RAN intelligence as a coordinated hierarchy across the Non-Real-Time (Non-RT), Near-Real-Time (Near-RT), and Real-Time (RT) control loops: (i) A Large Language Model (LLM) agent in the Non-RT RIC translates operator intent into policies and governs model lifecycles. (ii) Small Language Model (SLM) agents in the Near-RT RIC execute low-latency optimization and can activate, tune, or disable existing control applications; and (iii) Wireless Physical-layer Foundation Model (WPFM) agents near the distributed unit provide fast inference close to the air interface. We describe how these agents cooperate through standardized O-RAN interfaces and telemetry. Using a proof-of-concept implementation built on open-source models, software, and datasets, we demonstrate the proposed agentic approach in two representative scenarios: robust operation under non-stationary conditions and intent-driven slice resource control.

Sparse Autoencoders (SAEs) have emerged as a promising tool for interpreting neural networks by decomposing their activations into sparse sets of human-interpretable features. Recent work has introduced multiple SAE variants and successfully scaled them to frontier models. Despite much excitement, a growing number of negative results in downstream tasks casts doubt on whether SAEs recover meaningful features. To directly investigate this, we perform two complementary evaluations. On a synthetic setup with known ground-truth features, we demonstrate that SAEs recover only $9\%$ of true features despite achieving $71\%$ explained variance, showing that they fail at their core task even when reconstruction is strong. To evaluate SAEs on real activations, we introduce three baselines that constrain SAE feature directions or their activation patterns to random values. Through extensive experiments across multiple SAE architectures, we show that our baselines match fully-trained SAEs in interpretability (0.87 vs 0.90), sparse probing (0.69 vs 0.72), and causal editing (0.73 vs 0.72). Together, these results suggest that SAEs in their current state do not reliably decompose models' internal mechanisms.

Predicting flows that occur both through and around porous bodies is challenging due to coupled physics across fluid and porous regions and the need to generalize across diverse geometries and boundary conditions. We address this problem using two Physics Informed learning approaches: Physics Informed PointNets (PIPN) and Physics Informed Geometry Aware Neural Operator (P-IGANO). We enforce the incompressible Navier Stokes equations in the free-flow region and a Darcy Forchheimer extension in the porous region within a unified loss and condition the networks on geometry and material parameters. Datasets are generated with OpenFOAM on 2D ducts containing porous obstacles and on 3D windbreak scenarios with tree canopies and buildings. We first verify the pipeline via the method of manufactured solutions, then assess generalization to unseen shapes, and for PI-GANO, to variable boundary conditions and parameter settings. The results show consistently low velocity and pressure errors in both seen and unseen cases, with accurate reproduction of the wake structures. Performance degrades primarily near sharp interfaces and in regions with large gradients. Overall, the study provides a first systematic evaluation of PIPN/PI-GANO for simultaneous through-and-around porous flows and shows their potential to accelerate design studies without retraining per geometry.

Edge-cloud synergies provide a promising paradigm for privacy-preserving deployment of foundation models, where lightweight on-device models adapt to domain-specific data and cloud-hosted models coordinate knowledge sharing. However, in real-world edge environments, collaborative multimodal learning is challenged by modality heterogeneity (different modality combinations across domains) and model-structure heterogeneity (different modality-specific encoders/fusion modules. To address these issues, we propose ML-ECS, a collaborative multimodal learning framework that enables joint training between a server-based model and heterogeneous edge models. This framework consists of four components: (1) cross-modal contrastive learning (CCL) to align modality representations in a shared latent space, (2) adaptive multimodal tuning (AMT) to preserve domain-specific knowledge from local datasets, (3) modality-aware model aggregation (MMA) to robustly aggregate while mitigating noise caused by missing modalities, and (4) SLM-enhanced CCL (SE-CCL) to facilitate bidirectional knowledge transfer between cloud and edge. Experimental results on various multimodal tasks show that \pname consistently outperform state-of-the-art baselines under varying modality availability, achieving improvements of 5.44% to 12.08% in Rouge-LSum and improving both client- and server-side performance. In addition, by communicating only low-rank LoRA parameters and fused representations, ML-ECS achieves high communication efficiency, requiring only 0.65% of the total parameter volume.

The most valuable asset of any cloud-based organization is data, which is increasingly exposed to sophisticated cyberattacks. Until recently, the implementation of security measures in DevOps environments was often considered optional by many government entities and critical national services operating in the cloud. This includes systems managing sensitive information, such as electoral processes or military operations, which have historically been valuable targets for cybercriminals. Resistance to security implementation is often driven by concerns over losing agility in software development, increasing the risk of accumulated vulnerabilities. Nowadays, patching software is no longer enough; adopting a proactive cyber defense strategy, supported by Artificial Intelligence (AI), is crucial to anticipating and mitigating threats. Thus, this work proposes integrating the Security Chaos Engineering (SCE) methodology with a new LLM-based flow to automate the creation of attack defense trees that represent adversary behavior and facilitate the construction of SCE experiments based on these graphical models, enabling teams to stay one step ahead of attackers and implement previously unconsidered defenses. Further detailed information about the experiment performed, along with the steps to replicate it, can be found in the following repository: https://github.com/mariomc14/devsecops-adversary-llm.git.

Whether neural networks can serve as cognitive models of morphological learning remains an open question. Recent work has shown that encoder-decoder models can acquire irregular patterns, but evidence that they generalize these patterns like humans is mixed. We investigate this using the Spanish \emph{L-shaped morphome}, where only the first-person singular indicative (e.g., \textit{pongo} `I put') shares its stem with all subjunctive forms (e.g., \textit{ponga, pongas}) despite lacking apparent phonological, semantic, or syntactic motivation. We compare five encoder-decoder transformers varying along two dimensions: sequential vs. position-invariant positional encoding, and atomic vs. decomposed tag representations. Positional encoding proves decisive: position-invariant models recover the correct L-shaped paradigm clustering even when L-shaped verbs are scarce in training, whereas sequential positional encoding models only partially capture the pattern. Yet none of the models productively generalize this pattern to novel forms. Position-invariant models generalize the L-shaped stem across subjunctive cells but fail to extend it to the first-person singular indicative, producing a mood-based generalization rather than the L-shaped morphomic pattern. Humans do the opposite, generalizing preferentially to the first-person singular indicative over subjunctive forms. None of the models reproduce the human pattern, highlighting the gap between statistical pattern reproduction and morphological abstraction.

As robots become increasingly integrated into everyday tasks, their ability to perceive both the shape and properties of objects during in-hand manipulation becomes critical for adaptive and intelligent behavior. We present SemanticFeels, an extension of the NeuralFeels framework that integrates semantic labeling with neural implicit shape representation, from vision and touch. To illustrate its application, we focus on material classification: high-resolution Digit tactile readings are processed by a fine-tuned EfficientNet-B0 convolutional neural network (CNN) to generate local material predictions, which are then embedded into an augmented signed distance field (SDF) network that jointly predicts geometry and continuous material regions. Experimental results show that the system achieves a high correspondence between predicted and actual materials on both single- and multi-material objects, with an average matching accuracy of 79.87% across multiple manipulation trials on a multi-material object.

Monitoring chain-of-thought (CoT) reasoning is a foundational safety technique for large language model (LLM) agents; however, this oversight is compromised if models learn to conceal their reasoning. We explore the potential for steganographic CoT -- where models hide secret reasoning within innocuous text -- to inform risk assessment and deployment policies. We systematically evaluate the limits of steganographic capabilities across 28 models, ranging from past generations to the current frontier. We measure monitor evasion, refusal rates, encoding fidelity, and hidden task accuracy across four datasets, comparing steganographic acrostics against plain reasoning and filler-token baselines. We find that current models cannot yet sustain hidden reasoning for complex math and arithmetic tasks. However, in a simplified counting experiment, Claude Opus 4.5 achieved 92% accuracy on the hidden task, demonstrating nascent capability. Notably, in rare cases (<1%), GPT-5.2 might refuse steganographic instructions while simultaneously complying with them. Our findings underscore the need for continuous evaluation of steganographic risks. This study provides a methodology to preemptively detect and prevent hidden reasoning that might empower misaligned scheming and deceptive behavior.

Post-training GUI agents in interactive environments is critical for developing generalization and long-horizon planning capabilities. However, training on real-world applications is hindered by high latency, poor reproducibility, and unverifiable rewards relying on noisy visual proxies. To address the limitations, we present GUI-GENESIS, the first framework to automatically synthesize efficient GUI training environments with verifiable rewards. GUI-GENESIS reconstructs real-world applications into lightweight web environments using multimodal code models and equips them with code-native rewards, executable assertions that provide deterministic reward signals and eliminate visual estimation noise. Extensive experiments show that GUI-GENESIS reduces environment latency by 10 times and costs by over $28,000 per epoch compared to training on real applications. Notably, agents trained with GUI-GENESIS outperform the base model by 14.54% and even real-world RL baselines by 3.27% on held-out real-world tasks. Finally, we observe that models can synthesize environments they cannot yet solve, highlighting a pathway for self-improving agents.

Large language models (LLMs) have emerged as powerful tools for natural language table reasoning, where there are two main categories of methods. Prompt-based approaches rely on language-only inference or one-pass program generation without step-level verification. Agent-based approaches use tools in a closed loop, but verification is often local and backtracking is limited, allowing errors to propagate and increasing cost. Moreover, they rely on chain- or beam-style trajectories that are typically combinatorially redundant, leading to high token costs. In this paper, we propose TabTracer, an agentic framework that coordinates multi-step tool calls over intermediate table states, with explicit state tracking for verification and rollback. First, it enforces step-level verification with typed operations and lightweight numeric and format checks to provide reliable rewards and suppress hallucinations. Second, execution-feedback Monte Carlo Tree Search maintains a search tree of candidate table states and uses backpropagated reflection scores to guide UCB1 selection and rollback via versioned snapshots. Third, it reduces redundancy with budget-aware pruning, deduplication, and state hashing with a monotonicity gate to cut token cost. Comprehensive evaluation on TabFact, WikiTQ, and CRT datasets shows that TabTracer outperforms state-of-the-art baselines by up to 6.7% in accuracy while reducing token consumption by 59--84%.

We study Neural Optimal Transport in infinite-dimensional Hilbert spaces. In non-regular settings, Semi-dual Neural OT often generates spurious solutions that fail to accurately capture target distributions. We analytically characterize this spurious solution problem using the framework of regular measures, which generalize Lebesgue absolute continuity in finite dimensions. To resolve ill-posedness, we extend the semi-dual framework via a Gaussian smoothing strategy based on Brownian motion. Our primary theoretical contribution proves that under a regular source measure, the formulation is well-posed and recovers a unique Monge map. Furthermore, we establish a sharp characterization for the regularity of smoothed measures, proving that the success of smoothing depends strictly on the kernel of the covariance operator. Empirical results on synthetic functional data and time-series datasets demonstrate that our approach effectively suppresses spurious solutions and outperforms existing baselines.

Large Language Models (LLMs) have empowered autonomous agents to handle complex web navigation tasks. While recent studies integrate tree search to enhance long-horizon reasoning, applying these algorithms in web navigation faces two critical challenges: sparse valid paths that lead to inefficient exploration, and a noisy context that dilutes accurate state perception. To address this, we introduce Plan-MCTS, a framework that reformulates web navigation by shifting exploration to a semantic Plan Space. By decoupling strategic planning from execution grounding, it transforms sparse action space into a Dense Plan Tree for efficient exploration, and distills noisy contexts into an Abstracted Semantic History for precise state awareness. To ensure efficiency and robustness, Plan-MCTS incorporates a Dual-Gating Reward to strictly validate both physical executability and strategic alignment and Structural Refinement for on-policy repair of failed subplans. Extensive experiments on WebArena demonstrate that Plan-MCTS achieves state-of-the-art performance, surpassing current approaches with higher task effectiveness and search efficiency.

The generation of classical Chinese \textit{Ci} poetry, a form demanding a sophisticated blend of structural rigidity, rhythmic harmony, and artistic quality, poses a significant challenge for large language models (LLMs). To systematically evaluate and advance this capability, we introduce \textbf{C}hinese \textbf{Ci}pai \textbf{V}ariants (\textbf{CCiV}), a benchmark designed to assess LLM-generated \textit{Ci} poetry across these three dimensions: structure, rhythm, and quality. Our evaluation of 17 LLMs on 30 \textit{Cipai} reveals two critical phenomena: models frequently generate valid but unexpected historical variants of a poetic form, and adherence to tonal patterns is substantially harder than structural rules. We further show that form-aware prompting can improve structural and tonal control for stronger models, while potentially degrading weaker ones. Finally, we observe weak and inconsistent alignment between formal correctness and literary quality in our sample. CCiV highlights the need for variant-aware evaluation and more holistic constrained creative generation methods.

Standard factuality evaluations of LLMs treat all errors alike, obscuring whether failures arise from missing knowledge (empty shelves) or from limited access to encoded facts (lost keys). We propose a behavioral framework that profiles factual knowledge at the level of facts rather than questions, characterizing each fact by whether it is encoded, and then by how accessible it is: cannot be recalled, can be directly recalled, or can only be recalled with inference-time computation (thinking). To support such profiling, we introduce WikiProfile, a new benchmark constructed via an automated pipeline with a prompted LLM grounded in web search. Across 4 million responses from 13 LLMs, we find that encoding is nearly saturated in frontier models on our benchmark, with GPT-5 and Gemini-3 encoding 95--98% of facts. However, recall remains a major bottleneck: many errors previously attributed to missing knowledge instead stem from failures to access it. These failures are systematic and disproportionately affect long-tail facts and reverse questions. Finally, we show that thinking improves recall and can recover a substantial fraction of failures, indicating that future gains may rely less on scaling and more on methods that improve how models utilize what they already encode.

Despite their success, large pretrained vision models remain vulnerable to catastrophic forgetting when adapted to new tasks in class-incremental settings. Parameter-efficient fine-tuning (PEFT) alleviates this by restricting trainable parameters, yet most approaches still rely on cross-entropy (CE) loss, a surrogate for the 0-1 loss, to learn from new data. We revisit this choice and revive the true objective (0-1 loss) through a reinforcement learning perspective. By formulating classification as a one-step Markov Decision Process, we derive an Expected Policy Gradient (EPG) method that directly minimizes misclassification error with a low-variance gradient estimation. Our analysis shows that CE can be interpreted as EPG with an additional sample-weighting mechanism: CE encourages exploration by emphasizing low-confidence samples, while EPG prioritizes high-confidence ones. Building on this insight, we propose adaptive entropy annealing (aEPG), a training strategy that transitions from exploratory (CE-like) to exploitative (EPG-like) learning. aEPG-based methods outperform CE-based methods across diverse benchmarks and with various PEFT modules. More broadly, we evaluate various entropy regularization methods and demonstrate that lower entropy of the output prediction distribution enhances adaptation in pretrained vision models.

Inference-time scaling (ITS) in latent reasoning models typically introduces stochasticity through heuristic perturbations, such as dropout or fixed Gaussian noise. While these methods increase trajectory diversity, their exploration behavior is not explicitly modeled and can be inefficient under finite sampling budgets. We observe that stronger perturbations do not necessarily translate into more effective candidate trajectories, as unguided noise may disrupt internal decision structure rather than steer it. To provide a more structured alternative, we model latent thought exploration as conditional sampling from learnable densities and instantiate this idea as a Gaussian Thought Sampler (GTS). GTS predicts context-dependent perturbation distributions over continuous reasoning states and is trained with GRPO-style policy optimization while keeping the backbone frozen. Experiments on GSM8K with two latent reasoning architectures show that GTS achieves more reliable inference-time scaling than heuristic baselines. These findings indicate that improving latent ITS requires structured and optimizable exploration mechanisms rather than simply amplifying stochasticity.

Suppose you run a home exam, where students should report their own scores but can cheat freely. You can, if needed, call a limited number of students to class and verify their actual performance against their reported score. We consider the class of mechanisms where truthful reporting is a dominant strategy, and truthful agents are never penalized -- even off-equilibrium. How many students do we need to verify, in expectation, if we want to minimize the bias, i.e., the difference between agents' competence and their expected grade? When perfect verification is available, we characterize the best possible tradeoff between these requirements and provide a simple parametrized mechanism that is optimal in the class for any distribution of agents' types. When verification is noisy, the task becomes much more challenging. We show how proper scoring rules can be leveraged in different ways to construct truthful mechanisms with a good (though not necessarily optimal) tradeoff.

Most vision-language models (VLMs) are trained on English-centric data, limiting their performance in other languages and cultural contexts. This restricts their usability for non-English-speaking users and hinders the development of multimodal systems that reflect diverse linguistic and cultural realities. In this work, we reproduce and adapt the LLaVA-Next methodology to create a set of Polish VLMs. We rely on a fully automated pipeline for translating and filtering existing multimodal datasets, and complement this with synthetic Polish data for OCR and culturally specific tasks. Despite relying almost entirely on automatic translation and minimal manual intervention to the training data, our approach yields strong results: we observe a +9.5% improvement over LLaVA-1.6-Vicuna-13B on a Polish-adapted MMBench, along with higher-quality captions in generative evaluations, as measured by human annotators in terms of linguistic correctness. These findings highlight that large-scale automated translation, combined with lightweight filtering, can effectively bootstrap high-quality multimodal models for low-resource languages. Some challenges remain, particularly in cultural coverage and evaluation. To facilitate further research, we make our models and evaluation dataset publicly available.

Scalar reward models compress multi-dimensional human preferences into a single opaque score, creating an information bottleneck that often leads to brittleness and reward hacking in open-ended alignment. We argue that robust alignment for non-verifiable tasks is fundamentally a principle generalization problem: reward should not be a learned function internalized into a judge, but an explicit reasoning process executed under inspectable principles. To operationalize this view, we present the Open Rubric System (OpenRS), a plug-and-play, rubrics-based LLM-as-a-Judge framework built around Pairwise Adaptive Meta-Rubrics (PAMR) and lightweight Pointwise Verifiable Rubrics (PVRs), which provide both hard-constraint guardrails and verifiable reward components when ground-truth or programmatic checks are available. OpenRS uses an explicit meta-rubric -- a constitution-like specification that governs how rubrics are instantiated, weighted, and enforced -- and instantiates adaptive rubrics on the fly by conditioning on the semantic differences between two candidate responses. It then performs criterion-wise pairwise comparisons and aggregates criterion-level preferences externally, avoiding pointwise weighted scalarization while improving discriminability in open-ended settings. To keep principles consistent yet editable across various domains, we introduce a two-level meta-rubric refinement pipeline (automated evolutionary refinement for general principles and a reproducible human-in-the-loop procedure for domain principles), complemented with pointwise verifiable rubrics that act as both guardrails against degenerate behaviors and a source of verifiable reward for objective sub-tasks. Finally, we instantiate OpenRS as reward supervision in pairwise RL training.

Knowledge-intensive Visual Question Answering (KI-VQA) frequently suffers from severe knowledge conflicts caused by the inherent limitations of open-domain retrieval. However, existing paradigms face critical limitations due to the lack of generalizable conflict detection and intra-model constraint mechanisms to handle conflicting evidence. To address these challenges, we propose the REAL (Reasoning-Pivot Alignment) framework centered on the novel concept of the Reasoning-Pivot. Distinct from reasoning steps that prioritize internal self-derivation, a reasoning-pivot serves as an atomic unit (node or edge) in the reasoning chain that emphasizes knowledge linkage, and it typically relies on external evidence to complete the reasoning. Supported by our constructed REAL-VQA dataset, our approach integrates Reasoning-Pivot Aware SFT (RPA-SFT) to train a generalizable discriminator by aligning conflicts with pivot extraction, and employs Reasoning-Pivot Guided Decoding (RPGD), an intra-model decoding strategy that leverages these pivots for targeted conflict mitigation. Extensive experiments across diverse benchmarks demonstrate that REAL significantly enhances discrimination accuracy and achieves state-of-the-art performance, validating the effectiveness of our pivot-driven resolution paradigm.

Large, openly licensed speech datasets are essential for building automatic speech recognition (ASR) systems, yet many widely spoken languages remain underrepresented in public resources. Pashto, spoken by more than 60 million people, has historically lacked large-scale openly licensed speech data suitable for modern ASR development. This paper presents a release-level analysis of the Pashto component of the Mozilla Common Voice corpus, focusing on version 24.0 (December 2025) and contextualizing trends across major releases. We document rapid growth from 1.49 recorded hours in mid-2023 to 2,768.7 total hours in 2025, including 975.89 validated hours available for supervised ASR training. Beyond scale, we analyze validation throughput, contributor participation inequality, demographic metadata completeness, and sentence-level concentration in the validated subset. We find that participation is extremely concentrated (Gini = 0.941), age representation is strongly skewed toward young adults, and 41.97\% of clips lack self-reported gender labels, limiting subgroup auditing based on metadata. At the textual level, prompt reuse is moderate: 35.88\% of unique sentences account for 50\% of validated clips, suggesting that structural concentration is driven primarily by uneven contributor activity rather than dominance of a small prompt set. These results provide a quantitative audit of a rapidly scaling low-resource speech corpus and highlight practical priorities for improving dataset maturity, including expanded validation capacity and broader demographic participation.

We introduce LM-Lexicon, an innovative definition modeling approach that incorporates data clustering, semantic expert learning, and model merging using a sparse mixture-of-experts architecture. By decomposing the definition modeling task into specialized semantic domains, where small language models are trained as domain experts, LM-Lexicon achieves substantial improvements (+7% BLEU score compared with the prior state-of-the-art model) over existing methods on five widely used benchmarks. Empirically, we demonstrate that 1) the clustering strategy enables fine-grained expert specialization with nearly 10% improvement in definition quality; 2) the semantic-aware domain-level routing mechanism achieves higher expert efficacy (+1%) than conventional token-level routing; and 3) further performance gains can be obtained through test-time compute and semantic expert scaling. Our work advances definition modeling while providing insights into the development of efficient language models for semantic-intensive applications.

Code generation remains a challenging task that requires precise and structured reasoning. Existing Test Time Scaling (TTS) methods, including structured tree search, have made progress in exploring reasoning paths but still face two major challenges: (1) underthinking, where reasoning chains tend to be shallow and fail to capture the full complexity of problems; and (2) overthinking, where overly verbose reasoning leads to inefficiency and increased computational costs. To address these issues, we propose LogitsCoder, a novel framework that enhances chain-of-thought reasoning through lightweight, logit-level control mechanisms for code generation. LogitsCoder iteratively generates and refines reasoning steps by first steering token selection toward statistically preferred patterns via Logits Preference Decoding, then selecting and aggregating diverse reasoning paths using Logits Rank Based Path Selection and Thoughts Aggregation. This results in coherent and effective reasoning chains that balance depth and efficiency. Extensive experiments demonstrate that LogitsCoder produces more efficient and higher-quality reasoning chains, leading to superior code generation performance compared to baseline methods.

Decentralized federated learning (DFL) enables edge devices to collaboratively train models through local training and fully decentralized device-to-device (D2D) model exchanges. However, these energy-intensive operations often rapidly deplete limited device batteries, reducing their operational lifetime and degrading the learning performance. To address this limitation, we apply energy harvesting technique to DFL systems, allowing edge devices to extract ambient energy and operate sustainably. We first derive the convergence bound for wireless DFL with energy harvesting, showing that the convergence is influenced by partial device participation and transmission packet drops, both of which further depend on the available energy supply. To accelerate convergence, we formulate a joint device scheduling and power control problem and model it as a multi-agent Markov decision process (MDP). Traditional MDP algorithms (e.g., value or policy iteration) require a centralized coordinator with access to all device states and exhibit exponential complexity in the number of devices, making them impractical for large-scale decentralized networks. To overcome these challenges, we propose a fully decentralized policy iteration algorithm that leverages only local state information from two-hop neighboring devices, thereby substantially reducing both communication overhead and computational complexity. We further provide a theoretical analysis showing that the proposed decentralized algorithm achieves asymptotic optimality. Finally, comprehensive numerical experiments on real-world datasets are conducted to validate the theoretical results and corroborate the effectiveness of the proposed algorithm.

Length generalization is the ability of language models to maintain performance on inputs longer than those seen during pretraining. In this work, we introduce a simple yet powerful position encoding (PE) strategy, Random Float Sampling (RFS), that generalizes well to lengths unseen during pretraining or fine-tuning. In particular, instead of selecting position indices from a predefined discrete set, RFS uses randomly sampled continuous values, thereby avoiding out-of-distribution (OOD) issues on unseen lengths by exposing the model to diverse indices during training. Since assigning indices to tokens is a common and fundamental procedure in widely used PEs, the advantage of RFS can easily be incorporated into, for instance, the absolute sinusoidal encoding, RoPE, and ALiBi. Experiments corroborate its effectiveness by showing that RFS results in superior performance in length generalization tasks as well as zero-shot commonsense reasoning benchmarks.

Spatio-temporal traffic forecasting is a core component of intelligent transportation systems, supporting various downstream tasks such as signal control and network-level traffic management. In real-world deployments, forecasting models must operate under structural and observational uncertainties, conditions that are rarely considered in model design. Recent approaches achieve strong short-term predictive performance by tightly coupling spatial and temporal modeling, often at the cost of increased complexity and limited modularity. In contrast, efficient time-series models capture long-range temporal dependencies without relying on explicit network structure. We propose UniST-Pred, a unified spatio-temporal forecasting framework that first decouples temporal modeling from spatial representation learning, then integrates both through adaptive representation-level fusion. To assess robustness of the proposed approach, we construct a dataset based on an agent-based, microscopic traffic simulator (MATSim) and evaluate UniST-Pred under severe network disconnection scenarios. Additionally, we benchmark UniST-Pred on standard traffic prediction datasets, demonstrating its competitive performance against existing well-established models despite a lightweight design. The results illustrate that UniST-Pred maintains strong predictive performance across both real-world and simulated datasets, while also yielding interpretable spatio-temporal representations under infrastructure disruptions. The source code and the generated dataset are available at https://anonymous.4open.science/r/UniST-Pred-EF27

Maintenance is a critical stage in the software lifecycle, ensuring that post-release systems remain reliable, efficient, and adaptable. However, manual software maintenance is labor-intensive, time-consuming, and error-prone, which highlights the urgent need for automation. Learning from maintenance activities conducted on other software systems offers an effective way to improve efficiency. In particular, recent research has demonstrated that migration-based approaches transfer knowledge, artifacts, or solutions from one system to another and show strong potential in tasks such as API evolution adaptation, software testing, and migrating patches for fault correction. This makes migration-based maintenance a valuable research direction for advancing automated maintenance. This paper takes a step further by presenting the first systematic research agenda on migration-based approaches to software maintenance. We characterize the migration-based maintenance lifecycle through four key stages: \ding{182} identifying a maintenance task that can be addressed through migration, \ding{183} selecting suitable migration sources for the target project,\ding{184} matching relevant data across systems and adapting the migrated data to the target context, and \ding{185} validating the correctness of the migration. We also analyze the challenges that may arise at each stage. Our goal is to encourage the community to explore migration-based approaches more thoroughly and to tackle the key challenges that must be solved to advance automated software maintenance.

Large language models (LLMs) show strong reasoning abilities across diverse tasks, yet their performance on extended contexts remains inconsistent. While prior research has emphasized mid-context degradation in question answering, this study examines the impact of context in LLM-based fact verification. Using three datasets (HOVER, FEVEROUS, and ClimateFEVER) and five open-source models accross different parameters sizes (7B, 32B and 70B parameters) and model families (Llama-3.1, Qwen2.5 and Qwen3), we evaluate both parametric factual knowledge and the impact of evidence placement across varying context lengths. We find that LLMs exhibit non-trivial parametric knowledge of factual claims and that their verification accuracy generally declines as context length increases. Similarly to what has been shown in previous works, in-context evidence placement plays a critical role with accuracy being consistently higher when relevant evidence appears near the beginning or end of the prompt and lower when placed mid-context. These results underscore the importance of prompt structure in retrieval-augmented fact-checking systems.

Social simulation is critical for mining complex social dynamics and supporting data-driven decision making. LLM-based methods have emerged as powerful tools for this task by leveraging human-like social questionnaire responses to model group behaviors. Existing LLM-based approaches predominantly focus on group-level values at discrete time points, treating them as static snapshots rather than dynamic processes. However, group-level values are not fixed but shaped by long-term social changes. Modeling their dynamics is thus crucial for accurate social evolution prediction--a key challenge in both data mining and social science. This problem remains underexplored due to limited longitudinal data, group heterogeneity, and intricate historical event impacts. To bridge this gap, we propose a novel framework for group-level dynamic social simulation by integrating historical value trajectories into LLM-based human response modeling. We select China and the U.S. as representative contexts, conducting stratified simulations across four core sociodemographic dimensions (gender, age, education, income). Using the World Values Survey, we construct a multi-wave, group-level longitudinal dataset to capture historical value evolution, and then propose the first event-based prediction method for this task, unifying social events, current value states, and group attributes into a single framework. Evaluations across five LLM families show substantial gains: a maximum 30.88\% improvement on seen questions and 33.97\% on unseen questions over the Vanilla baseline. We further find notable cross-group heterogeneity: U.S. groups are more volatile than Chinese groups, and younger groups in both countries are more sensitive to external changes. These findings advance LLM-based social simulation and provide new insights for social scientists to understand and predict social value changes.

In real-world scenarios, the performance of semantic segmentation often deteriorates when processing low-quality (LQ) images, which may lack clear semantic structures and high-frequency details. Although image restoration techniques offer a promising direction for enhancing degraded visual content, conventional real-world image restoration (Real-IR) models primarily focus on pixel-level fidelity and often fail to recover task-relevant semantic cues, limiting their effectiveness when directly applied to downstream vision tasks. Conversely, existing segmentation models trained on high-quality data lack robustness under real-world degradations. In this paper, we propose Restoration Adaptation for Semantic Segmentation (RASS), which effectively integrates semantic image restoration into the segmentation process, enabling high-quality semantic segmentation on the LQ images directly. Specifically, we first propose a Semantic-Constrained Restoration (SCR) model, which injects segmentation priors into the restoration model by aligning its cross-attention maps with segmentation masks, encouraging semantically faithful image reconstruction. Then, RASS transfers semantic restoration knowledge into segmentation through LoRA-based module merging and task-specific fine-tuning, thereby enhancing the model's robustness to LQ images. To validate the effectiveness of our framework, we construct a real-world LQ image segmentation dataset with high-quality annotations, and conduct extensive experiments on both synthetic and real-world LQ benchmarks. The results show that SCR and RASS significantly outperform state-of-the-art methods in segmentation and restoration tasks. Code, models, and datasets will be available at https://github.com/Ka1Guan/RASS.git.

We present BitDance, a scalable autoregressive (AR) image generator that predicts binary visual tokens instead of codebook indices. With high-entropy binary latents, BitDance lets each token represent up to $2^{256}$ states, yielding a compact yet highly expressive discrete representation. Sampling from such a huge token space is difficult with standard classification. To resolve this, BitDance uses a binary diffusion head: instead of predicting an index with softmax, it employs continuous-space diffusion to generate the binary tokens. Furthermore, we propose next-patch diffusion, a new decoding method that predicts multiple tokens in parallel with high accuracy, greatly speeding up inference. On ImageNet 256x256, BitDance achieves an FID of 1.24, the best among AR models. With next-patch diffusion, BitDance beats state-of-the-art parallel AR models that use 1.4B parameters, while using 5.4x fewer parameters (260M) and achieving 8.7x speedup. For text-to-image generation, BitDance trains on large-scale multimodal tokens and generates high-resolution, photorealistic images efficiently, showing strong performance and favorable scaling. When generating 1024x1024 images, BitDance achieves a speedup of over 30x compared to prior AR models. We release code and models to facilitate further research on AR foundation models. Code and models are available at: https://github.com/shallowdream204/BitDance.

Mixture-of-Experts (MoE) embedding models combine expert outputs using weighted linear summation, implicitly assuming a linear subspace structure in the embedding space. This assumption is shown to be inconsistent with the geometry of expert representations. Geometric analysis of a modern MoE embedding model reveals that expert outputs lie on a shared hyperspherical manifold characterized by tightly concentrated norms and substantial angular separation. Under this geometry, linear aggregation induces inward collapse toward the manifold interior, distorting vector magnitude and direction and reducing embedding comparability. To address this inconsistency, Spherical Barycentric Aggregation (SBA) is introduced as a geometry-preserving aggregation operator that separates radial and angular components to maintain hyperspherical structure while remaining fully compatible with existing routing mechanisms. Experiments on selected tasks from the Massive Text Embedding Benchmark (MTEB), including semantic similarity, clustering, and duplicate question detection, demonstrate consistent performance improvements with identical training cost and full stability. Additional geometric analyses confirm that SBA prevents aggregation-induced collapse and preserves hyperspherical consistency, highlighting the importance of geometry-aware aggregation in MoE embedding architectures.

Memory is critical for enabling large language model (LLM) based agents to maintain coherent behavior over long-horizon interactions. However, existing agent memory systems suffer from two key gaps: they rely on a one-size-fits-all memory structure and do not model memory structure selection as a context-adaptive decision, limiting their ability to handle heterogeneous interaction patterns and resulting in suboptimal performance. We propose a unified framework, FluxMem, that enables adaptive memory organization for LLM agents. Our framework equips agents with multiple complementary memory structures. It explicitly learns to select among these structures based on interaction-level features, using offline supervision derived from downstream response quality and memory utilization. To support robust long-horizon memory evolution, we further introduce a three-level memory hierarchy and a Beta Mixture Model-based probabilistic gate for distribution-aware memory fusion, replacing brittle similarity thresholds. Experiments on two long-horizon benchmarks, PERSONAMEM and LoCoMo, demonstrate that our method achieves average improvements of 9.18% and 6.14%.

Flowchart-oriented dialogue (FOD) systems aim to guide users through multi-turn decision-making or operational procedures by following a domain-specific flowchart to achieve a task goal. In this work, we formalize flowchart reasoning in FOD as grounding user input to flowchart nodes at each dialogue turn while ensuring node transition is consistent with the correct flowchart path. Despite recent advances of LLMs in task-oriented dialogue systems, adapting them to FOD still faces two limitations: (1) LLMs lack an explicit mechanism to represent and reason over flowchart topology, and (2) they are prone to hallucinations, leading to unfaithful flowchart reasoning. To address these limitations, we propose FloCA, a zero-shot flowchart-oriented conversational agent. FloCA uses an LLM for intent understanding and response generation while delegating flowchart reasoning to an external tool that performs topology-constrained graph execution, ensuring faithful and logically consistent node transitions across dialogue turns. We further introduce an evaluation framework with an LLM-based user simulator and five new metrics covering reasoning accuracy and interaction efficiency. Extensive experiments on FLODIAL and PFDial datasets highlight the bottlenecks of existing LLM-based methods and demonstrate the superiority of FloCA. Our codes are available at https://github.com/Jinzi-Zou/FloCA-flowchart-reasoning.

Large language models now produce text indistinguishable from human writing, which increases the need for reliable provenance tracing. Multi-bit watermarking can embed identifiers into generated text, but existing methods struggle to keep both text quality and watermark strength while carrying long messages. We propose MC$^2$Mark, a distortion-free multi-bit watermarking framework designed for reliable embedding and decoding of long messages. Our key technical idea is Multi-Channel Colored Reweighting, which encodes bits through structured token reweighting while keeping the token distribution unbiased, together with Multi-Layer Sequential Reweighting to strengthen the watermark signal and an evidence-accumulation detector for message recovery. Experiments show that MC$^2$Mark improves detectability and robustness over prior multi-bit watermarking methods while preserving generation quality, achieving near-perfect accuracy for short messages and exceeding the second-best method by nearly 30% for long messages.

Iterative self-training (self-distillation) repeatedly refits a model on pseudo-labels generated by its own predictions. We study this procedure in overparameterized linear regression: an initial estimator is trained on noisy labels, and each subsequent iterate is trained on fresh covariates with noiseless pseudo-labels from the previous model. In the high-dimensional regime, we derive deterministic-equivalent recursions for the prediction risk and effective noise across iterations, and prove that the empirical quantities concentrate sharply around these limits. The recursion separates two competing forces: a systematic component that grows with iteration due to progressive signal forgetting, and a stochastic component that decays due to denoising via repeated data-dependent projections. Their interaction yields a $U$-shaped test-risk curve and an optimal early-stopping time. In spiked covariance models, iteration further acts as an iteration-dependent spectral filter that preserves strong eigendirections while suppressing weaker ones, inducing an implicit form of soft feature selection distinct from ridge regression. Finally, we propose an iterated generalized cross-validation criterion and prove its uniform consistency for estimating the risk along the self-training trajectory, enabling fully data-driven selection of the stopping time and regularization. Experiments on synthetic covariances validate the theory and illustrate the predicted denoising-forgetting trade-off.

While Group Relative Policy Optimization (GRPO) offers a powerful framework for LLM post-training, its effectiveness in open-ended domains like Machine Translation hinges on accurate intra-group ranking. We identify that standard Scalar Quality Metrics (SQM) fall short in this context; by evaluating candidates in isolation, they lack the comparative context necessary to distinguish fine-grained linguistic nuances. To address this, we introduce the Group Quality Metric (GQM) paradigm and instantiate it via the Group Relative Reward Model (GRRM). Unlike traditional independent scorers, GRRM processes the entire candidate group jointly, leveraging comparative analysis to rigorously resolve relative quality and adaptive granularity. Empirical evaluations confirm that GRRM achieves competitive ranking accuracy among all baselines. Building on this foundation, we integrate GRRM into the GRPO training loop to optimize the translation policy. Experimental results demonstrate that our framework not only improves general translation quality but also unlocks reasoning capabilities comparable to state-of-the-art reasoning models. We release codes, datasets, and model checkpoints at https://github.com/NJUNLP/GRRM.

Most time series foundation models are pretrained by directly predicting future observations, which often yields weakly structured latent representations that capture surface noise rather than coherent and predictable temporal dynamics. In this work, we introduce EIDOS, a foundation model family that shifts pretraining from future value prediction to latent-space predictive learning. We train a causal Transformer to predict the evolution of latent representations, encouraging the emergence of structured and temporally coherent latent states. To ensure stable targets for latent-space learning, we design a lightweight aggregation branch to construct target representations. EIDOS is optimized via a joint objective that integrates latent-space alignment, observational grounding to anchor representations to the input signal, and direct forecasting supervision. On the GIFT-Eval benchmark, EIDOS mitigates structural fragmentation in the representation space and achieves state-of-the-art performance. These results demonstrate that constraining models to learn predictable latent dynamics is a principled step toward more robust and reliable time series foundation models.

We study operator-norm covariance estimation from heavy-tailed samples that may include a small fraction of arbitrary outliers. A simple and widely used safeguard is \emph{Euclidean norm clipping}, but its accuracy depends critically on an unknown clipping level. We propose a cross-fitted clipped covariance estimator equipped with \emph{fully computable} Bernstein-type deviation certificates, enabling principled data-driven tuning via a selector (\emph{MinUpper}) that balances certified stochastic error and a robust hold-out proxy for clipping bias. The resulting procedure adapts to intrinsic complexity measures such as effective rank under mild tail regularity and retains meaningful guarantees under only finite fourth moments. Experiments on contaminated spiked-covariance benchmarks illustrate stable performance and competitive accuracy across regimes.

Subseasonal-to-seasonal (S2S) forecasts play an essential role in providing a decision-critical weeks-to-months planning window for climate resilience and sustainability, yet a growing bottleneck is the last-mile gap: translating scientific forecasts into trusted, actionable climate services, requiring reliable multimodal understanding and decision-facing reasoning under uncertainty. Meanwhile, multimodal large language models (MLLMs) and corresponding agentic paradigms have made rapid progress in supporting various workflows, but it remains unclear whether they can reliably generate decision-making deliverables from operational service products (e.g., actionable signal comprehension, decision-making handoff, and decision analysis & planning) under uncertainty. We introduce S2SServiceBench, a multimodal benchmark for last-mile S2S climate services curated from an operational climate-service system to evaluate this capability. S2SServiceBenchcovers 10 service products with about 150+ expert-selected cases in total, spanning six application domains - Agriculture, Disasters, Energy, Finance, Health, and Shipping. Each case is instantiated at three service levels, yielding around 500 tasks and 1,000+ evaluation items across climate resilience and sustainability applications. Using S2SServiceBench, we benchmark state-of-the-art MLLMs and agents, and analyze performance across products and service levels, revealing persistent challenges in S2S service plot understanding and reasoning - namely, actionable signal comprehension, operationalizing uncertainty into executable handoffs, and stable, evidence-grounded analysis and planning for dynamic hazards-while offering actionable guidance for building future climate-service agents.

The integration of LLMs into vulnerability detection (VD) has shifted the field toward interpretable and context-aware analysis. While post-training methods have shown promise in general coding tasks, their systematic application to VD remains underexplored. In this paper, we present the first comprehensive investigation into the post-training pipeline for LLM-based VD, spanning from cold-start SFT to off-policy preference optimization and on-policy RL, uncovering how data curation, stage interactions, reward mechanisms, and evaluation protocols collectively dictate the efficacy of model training and assessment. Our study identifies practical guidelines and insights: (1) SFT based on rejection sampling greatly outperforms rationalization-based supervision, which can introduce hallucinations due to ground-truth leakage. (2) While increased SFT epochs constantly benefit preference optimization, excessive SFT inhibits self-exploration during RL, ultimately limiting performance gains. (3) Coarse-grained reward signals often mislead RL, whereas fine-grained root-cause judgments ensure reliable credit assignment. Specification-based rewards offer further benefits but incur significant effort in specification generation. (4) Although filtering extremely hard-to-detect vulnerability samples improves RL training efficiency, the cost of performance loss should be considered in practical applications. (5) Models trained under GRPO significantly outperform those using SFT and preference optimization (i.e., DPO and ORPO), as well as a series of zero-shot SOTA LLMs, underscoring the significant potential of on-policy RL for LLM-based VD. (6) In contrast to binary matching that tends to overestimate performance, LLM-as-a-Judge based on root-cause analysis provides a more robust evaluation protocol, although its accuracy varies across judge models with different levels of security expertise.

Representing and predicting high-dimensional and spatiotemporally chaotic dynamical systems remains a fundamental challenge in dynamical systems and machine learning. Although data-driven models can achieve accurate short-term forecasts, they often lack stability, interpretability, and scalability in regimes dominated by broadband or continuous spectra. Koopman-based approaches provide a principled linear perspective on nonlinear dynamics, but existing methods rely on restrictive finite-dimensional assumptions or explicit spectral parameterizations that degrade in high-dimensional settings. Against these issues, we introduce KoopGen, a generator-based neural Koopman framework that models dynamics through a structured, state-dependent representation of Koopman generators. By exploiting the intrinsic Cartesian decomposition into skew-adjoint and self-adjoint components, KoopGen separates conservative transport from irreversible dissipation while enforcing exact operator-theoretic constraints during learning. Across systems ranging from nonlinear oscillators to high-dimensional chaotic and spatiotemporal dynamics, KoopGen improves prediction accuracy and stability, while clarifying which components of continuous-spectrum dynamics admit interpretable and learnable representations.

Pathology foundation models (PFMs) have enabled robust generalization in computational pathology through large-scale datasets and expansive architectures, but their substantial computational cost, particularly for gigapixel whole slide images, limits clinical accessibility and scalability. Here, we present LitePath, a deployment-friendly foundational framework designed to mitigate model over-parameterization and patch level redundancy. LitePath integrates LiteFM, a compact model distilled from three large PFMs (Virchow2, H-Optimus-1 and UNI2) using 190 million patches, and the Adaptive Patch Selector (APS), a lightweight component for task-specific patch selection. The framework reduces model parameters by 28x and lowers FLOPs by 403.5x relative to Virchow2, enabling deployment on low-power edge hardware such as the NVIDIA Jetson Orin Nano Super. On this device, LitePath processes 208 slides per hour, 104.5x faster than Virchow2, and consumes 0.36 kWh per 3,000 slides, 171x lower than Virchow2 on an RTX3090 GPU. We validated accuracy using 37 cohorts across four organs and 26 tasks (26 internal, 9 external, and 2 prospective), comprising 15,672 slides from 9,808 patients disjoint from the pretraining data. LitePath ranks second among 19 evaluated models and outperforms larger models including H-Optimus-1, mSTAR, UNI2 and GPFM, while retaining 99.71% of the AUC of Virchow2 on average. To quantify the balance between accuracy and efficiency, we propose the Deployability Score (D-Score), defined as the weighted geometric mean of normalized AUC and normalized FLOP, where LitePath achieves the highest value, surpassing Virchow2 by 10.64%. These results demonstrate that LitePath enables rapid, cost-effective and energy-efficient pathology image analysis on accessible hardware while maintaining accuracy comparable to state-of-the-art PFMs and reducing the carbon footprint of AI deployment.

Named Entity Recognition (NER) has emerged as a critical component in automating financial transaction processing, particularly in extracting structured information from unstructured payment data. This paper presents a comprehensive analysis of state-of-the-art NER algorithms specifically designed for payment data extraction, including Conditional Random Fields (CRF), Bidirectional Long Short-Term Memory with CRF (BiLSTM-CRF), and transformer-based models such as BERT and FinBERT. We conduct extensive experiments on a dataset of 50,000 annotated payment transactions across multiple payment formats including SWIFT MT103, ISO 20022, and domestic payment systems. Our experimental results demonstrate that fine-tuned BERT models achieve an F1-score of 94.2% for entity extraction, outperforming traditional CRF-based approaches by 12.8 percentage points. Furthermore, we introduce PaymentBERT, a novel hybrid architecture combining domain-specific financial embeddings with contextual representations, achieving state-of-the-art performance with 95.7% F1-score while maintaining real-time processing capabilities. We provide detailed analysis of cross-format generalization, ablation studies, and deployment considerations. This research provides practical insights for financial institutions implementing automated sanctions screening, anti-money laundering (AML) compliance, and payment processing systems.

The large artificial intelligence models (LAMs) show strong capabilities in perception, reasoning, and multi-modal understanding, and can enable advanced capabilities in low-altitude edge intelligence. However, the deployment of LAMs at the edge remains constrained by some fundamental limitations. First, tasks are rigidly tied to specific models, limiting the flexibility. Besides, the computational and memory demands of full-scale LAMs exceed the capacity of most edge devices. Moreover, the current inference pipelines are typically static, making it difficult to respond to real-time changes of tasks. To address these challenges, we propose a prompt-to-agent edge cognition framework (P2AECF), enabling the flexible, efficient, and adaptive edge intelligence. Specifically, P2AECF transforms high-level semantic prompts into executable reasoning workflows through three key mechanisms. First, the prompt-defined cognition parses task intent into abstract and model-agnostic representations. Second, the agent-based modular execution instantiates these tasks using lightweight and reusable cognitive agents dynamically selected based on current resource conditions. Third, the diffusion-controlled inference planning adaptively constructs and refines execution strategies by incorporating runtime feedback and system context. In addition, we illustrate the framework through a representative low-altitude intelligent network use case, showing its ability to deliver adaptive, modular, and scalable edge intelligence for real-time low-altitude aerial collaborations.

Large Language Models increasingly rely on self-explanations, such as chain of thought reasoning, to improve performance on multi step question answering. While these explanations enhance accuracy, they are often verbose and costly to generate, raising the question of how much explanation is truly necessary. In this paper, we examine the trade-off between sufficiency, defined as the ability of an explanation to justify the correct answer, and conciseness, defined as the reduction in explanation length. Building on the information bottleneck principle, we conceptualize explanations as compressed representations that retain only the information essential for producing correct answers.To operationalize this view, we introduce an evaluation pipeline that constrains explanation length and assesses sufficiency using multiple language models on the ARC Challenge dataset. To broaden the scope, we conduct experiments in both English, using the original dataset, and Persian, as a resource-limited language through translation. Our experiments show that more concise explanations often remain sufficient, preserving accuracy while substantially reducing explanation length, whereas excessive compression leads to performance degradation.

The unit testing of Deep Learning (DL) libraries is challenging due to complex numerical semantics and implicit tensor constraints. Traditional Search-Based Software Testing (SBST) often suffers from semantic blindness, failing to satisfy the constraints of high-dimensional tensors, whereas Large Language Models (LLMs) struggle with cross-file context and unstable code modifications. This paper proposes ATTest, an agent-driven tensor testing framework for module-level unit test generation. ATTest orchestrates a seven-stage pipeline, which encompasses constraint extraction and an iterative "generation-validation-repair" loop, to maintain testing stability and mitigate context-window saturation. An evaluation on PyTorch and TensorFlow demonstrates that ATTest significantly outperforms state-of-the-art baselines such as PynguinML, achieving an average branch coverage of 55.60% and 54.77%, respectively. The results illustrate how agent-driven workflows bridge the semantic gap in numerical libraries while ensuring auditable test synthesis. Source code: https://github.com/iSEngLab/ATTest.git

Artificial intelligence (AI) has demonstrated strong potential in clinical diagnostics, often achieving accuracy comparable to or exceeding that of human experts. A key challenge, however, is that AI reasoning frequently diverges from structured clinical frameworks, limiting trust, interpretability, and adoption. Critical symptoms, pivotal for rapid and accurate decision-making, may be overlooked by AI models even when predictions are correct. Existing post hoc explanation methods provide limited transparency and lack formal guarantees. To address this, we leverage formal abductive explanations, which offer consistent, guaranteed reasoning over minimal sufficient feature sets. This enables a clear understanding of AI decision-making and allows alignment with clinical reasoning. Our approach preserves predictive accuracy while providing clinically actionable insights, establishing a robust framework for trustworthy AI in medical diagnosis.

Providing extensive context via prompting is vital for leveraging the capabilities of Large Language Models (LLMs). However, lengthy contexts significantly increase inference latency, as the computational cost of self-attention grows quadratically with sequence length. To mitigate this issue, context compression-particularly soft prompt compressio-has emerged as a widely studied solution, which converts long contexts into shorter memory embeddings via a trained compressor. Existing methods typically compress the entire context indiscriminately into a set of memory tokens, requiring the compressor to capture global dependencies and necessitating extensive pre-training data to learn effective patterns. Inspired by the chunking mechanism in human working memory and empirical observations of the spatial specialization of memory embeddings relative to original tokens, we propose Parallelized Iterative Compression (PIC). By simply modifying the Transformer's attention mask, PIC explicitly restricts the receptive field of memory tokens to sequential local chunks, thereby lowering the difficulty of compressor training. Experiments across multiple downstream tasks demonstrate that PIC consistently outperforms competitive baselines, with superiority being particularly pronounced in high compression scenarios (e.g., achieving relative improvements of 29.8\% in F1 score and 40.7\% in EM score on QA tasks at the $64\times$ compression ratio). Furthermore, PIC significantly expedites the training process. Specifically, when training the 16$\times$ compressor, it surpasses the peak performance of the competitive baseline while effectively reducing the training time by approximately 40\%.

Alzheimer's disease (AD) has become a prevalent neurodegenerative disease worldwide. Traditional diagnosis still relies heavily on medical imaging and clinical assessment by physicians, which is often time-consuming and resource-intensive in terms of both human expertise and healthcare resources. In recent years, large language models (LLMs) have been increasingly applied to the medical field using electronic health records (EHRs), yet their application in Alzheimer's disease assessment remains limited, particularly given that AD involves complex multifactorial etiologies that are difficult to observe directly through imaging modalities. In this work, we propose leveraging LLMs to perform Chain-of-Thought (CoT) reasoning on patients' clinical EHRs. Unlike direct fine-tuning of LLMs on EHR data for AD classification, our approach utilizes LLM-generated CoT reasoning paths to provide the model with explicit diagnostic rationale for AD assessment, followed by structured CoT-based predictions. This pipeline not only enhances the model's ability to diagnose intrinsically complex factors but also improves the interpretability of the prediction process across different stages of AD progression. Experimental results demonstrate that the proposed CoT-based diagnostic framework significantly enhances stability and diagnostic performance across multiple CDR grading tasks, achieving up to a 15% improvement in F1 score compared to the zero-shot baseline method.

Reinforcement learning (RL) promises to unlock capabilities beyond imitation learning for Vision-Language-Action (VLA) models, but its requirement for massive real-world interaction prevents direct deployment on physical robots. Recent work attempts to use learned world models as simulators for policy optimization, yet closed-loop imagined rollouts inevitably suffer from hallucination and long-horizon error accumulation. Such errors do not merely degrade visual fidelity; they corrupt the optimization signal, encouraging policies to exploit model inaccuracies rather than genuine task progress. We propose WoVR, a reliable world-model-based reinforcement learning framework for post-training VLA policies. Instead of assuming a faithful world model, WoVR explicitly regulates how RL interacts with imperfect imagined dynamics. It improves rollout stability through a controllable action-conditioned video world model, reshapes imagined interaction to reduce effective error depth via Keyframe-Initialized Rollouts, and maintains policy-simulator alignment through World Model-Policy co-evolution. Extensive experiments on LIBERO benchmarks and real-world robotic manipulation demonstrate that WoVR enables stable long-horizon imagined rollouts and effective policy optimization, improving average LIBERO success from 39.95% to 69.2% (+29.3 points) and real-robot success from 61.7% to 91.7% (+30.0 points). These results show that learned world models can serve as practical simulators for reinforcement learning when hallucination is explicitly controlled.

Recommendation systems are essential for personalizing e-commerce shopping experiences. Among these, Trigger-Induced Recommendation (TIR) has emerged as a key scenario, which utilizes a trigger item (explicitly represents a user's instantaneous interest), enabling precise, real-time recommendations. Although several trigger-based techniques have been proposed, most of them struggle to address the intent myopia issue, that is, a recommendation system overemphasizes the role of trigger items and narrowly focuses on suggesting commodities that are highly relevant to trigger items. Meanwhile, existing methods rely on collaborative behavior patterns between trigger and recommended items to identify the user's preferences, yet the sparsity of ID-based interaction restricts their effectiveness. To this end, we propose the Deep Adaptive Intent-Aware Network (DAIAN) that dynamically adapts to users' intent preferences. In general, we first extract the users' personalized intent representations by analyzing the correlation between a user's click and the trigger item, and accordingly retrieve the user's related historical behaviors to mine the user's diverse intent. Besides, sparse collaborative behaviors constrain the performance in capturing items associated with user intent. Hence, we reinforce similarity by leveraging a hybrid enhancer with ID and semantic information, followed by adaptive selection based on varying intents. Experimental results on public datasets and our industrial e-commerce datasets demonstrate the effectiveness of DAIAN.

Most evaluations of External Memory Module assume a static setting: memory is built offline and queried at a fixed state. In practice, memory is streaming: new facts arrive continuously, insertions interleave with retrievals, and the memory state evolves while the model is serving queries. In this regime, accuracy and cost are governed by the full memory lifecycle, which encompasses the ingestion, maintenance, retrieval, and integration of information into generation. We present Neuromem, a scalable testbed that benchmarks External Memory Modules under an interleaved insertion-and-retrieval protocol and decomposes its lifecycle into five dimensions including memory data structure, normalization strategy, consolidation policy, query formulation strategy, and context integration mechanism. Using three representative datasets LOCOMO, LONGMEMEVAL, and MEMORYAGENTBENCH, Neuromem evaluates interchangeable variants within a shared serving stack, reporting token-level F1 and insertion/retrieval latency. Overall, we observe that performance typically degrades as memory grows across rounds, and time-related queries remain the most challenging category. The memory data structure largely determines the attainable quality frontier, while aggressive compression and generative integration mechanisms mostly shift cost between insertion and retrieval with limited accuracy gain.

Humanity's Last Exam (HLE) has become a widely used benchmark for evaluating frontier large language models on challenging, multi-domain questions. However, community-led analyses have raised concerns that HLE contains a non-trivial number of noisy items, which can bias evaluation results and distort cross-model comparisons. To address this challenge, we introduce HLE-Verified, a verified and revised version of HLE with a transparent verification protocol and fine-grained error taxonomy. Our construction follows a two-stage validation-and-repair workflow resulting in a certified benchmark. In Stage I, each item undergoes binary validation of the problem and final answer through domain-expert review and model-based cross-checks, yielding 641 verified items. In Stage II, flawed but fixable items are revised under strict constraints preserving the original evaluation intent, through dual independent expert repairs, model-assisted auditing, and final adjudication, resulting in 1,170 revised-and-certified items. The remaining 689 items are released as a documented uncertain set with explicit uncertainty sources and expertise tags for future refinement. We evaluate seven state-of-the-art language models on HLE and HLE-Verified, observing an average absolute accuracy gain of 7--10 percentage points on HLE-Verified. The improvement is particularly pronounced on items where the original problem statement and/or reference answer is erroneous, with gains of 30--40 percentage points. Our analyses further reveal a strong association between model confidence and the presence of errors in the problem statement or reference answer, supporting the effectiveness of our revisions. Overall, HLE-Verified improves HLE-style evaluations by reducing annotation noise and enabling more faithful measurement of model capabilities. Data is available at: https://github.com/SKYLENAGE-AI/HLE-Verified

In modern software development, developers frequently need to understand code behavior at a glance -- whether reviewing pull requests, debugging issues, or navigating unfamiliar codebases. This ability to reason about dynamic program behavior is fundamental to effective software engineering and increasingly supported by Large Language Models (LLMs). However, existing studies on code reasoning focus primarily on isolated code snippets, overlooking the complexity of real-world scenarios involving external API interactions and unfamiliar functions. This gap hinders our understanding of what truly makes code reasoning challenging for LLMs across diverse programming contexts. We present CodeGlance, a multi-dimensional benchmark investigating code reasoning challenges across three realistic scenarios: intrinsic logic reasoning, API interaction reasoning, and unseen function reasoning. Through systematic evaluation of 7 state-of-the-art LLMs, we reveal that unseen function reasoning poses significant challenges especially for smaller models, with Qwen2.5-3b achieving only 6.0\% accuracy on unseen functions compared to 37.5\% on familiar APIs. We identify critical code complexity features -- including execution trace length, API invocation count, and control flow complexity -- that significantly impact code reasoning difficulty across scenarios. We further investigate how common augmentation strategies, including CoT, document retrieval, and code search, can improve reasoning performance, finding that their effectiveness varies substantially depending on whether challenges stem from logical complexity or knowledge gaps. These findings provide actionable guidance for developing more capable code reasoning systems and deploying LLM-based programming assistants in real-world software development.

Data-driven approaches like deep learning are rapidly advancing planetary science, particularly in Mars exploration. Despite recent progress, most existing benchmarks remain confined to closed-set supervised visual tasks and do not support text-guided retrieval for geospatial discovery. We introduce MarsRetrieval, a retrieval benchmark for evaluating vision-language models for Martian geospatial discovery. MarsRetrieval includes three tasks: (1) paired image-text retrieval, (2) landform retrieval, and (3) global geo-localization, covering multiple spatial scales and diverse geomorphic origins. We propose a unified retrieval-centric protocol to benchmark multimodal embedding architectures, including contrastive dual-tower encoders and generative vision-language models. Our evaluation shows MarsRetrieval is challenging: even strong foundation models often fail to capture domain-specific geomorphic distinctions. We further show that domain-specific fine-tuning is critical for generalizable geospatial discovery in planetary settings. Our code is available at https://github.com/ml-stat-Sustech/MarsRetrieval

Constant-stepsize stochastic approximation (SA) is widely used in learning for computational efficiency. For a fixed stepsize, the iterates typically admit a stationary distribution that is rarely tractable. Prior work shows that as the stepsize $α\downarrow 0$, the centered-and-scaled steady state converges weakly to a Gaussian random vector. However, for fixed $α$, this weak convergence offers no usable error bound for approximating the steady-state by its Gaussian limit. This paper provides explicit, non-asymptotic error bounds for fixed $α$. We first prove general-purpose theorems that bound the Wasserstein distance between the centered-scaled steady state and an appropriate Gaussian distribution, under regularity conditions for drift and moment conditions for noise. To ensure broad applicability, we cover both i.i.d. and Markovian noise models. We then instantiate these theorems for three representative SA settings: (1) stochastic gradient descent (SGD) for smooth strongly convex objectives, (2) linear SA, and (3) contractive nonlinear SA. We obtain dimension- and stepsize-dependent, explicit bounds in Wasserstein distance of order $α^{1/2}\log(1/α)$ for small $α$. Building on the Wasserstein approximation error, we further derive non-uniform Berry--Esseen-type tail bounds that compare the steady-state tail probability to Gaussian tails. We achieve an explicit error term that decays in both the deviation level and stepsize $α$. We adapt the same analysis for SGD beyond strongly convexity and study general convex objectives. We identify a non-Gaussian (Gibbs) limiting law under the correct scaling, which is validated numerically, and provide a corresponding pre-limit Wasserstein error bound.

Language models are widely used in chemistry for molecular property prediction and small-molecule generation, yet Natural Products (NPs) remain underexplored despite their importance in drug discovery. To address this gap, we develop NP-specific chemical language models (NPCLMs) by pre-training state-space models (Mamba and Mamba-2) and comparing them with transformer baselines (GPT). Using a dataset of about 1M NPs, we present the first systematic comparison of selective state-space models and transformers for NP-focused tasks, together with eight tokenization strategies including character-level, Atom-in-SMILES (AIS), byte-pair encoding (BPE), and NP-specific BPE. We evaluate molecule generation (validity, uniqueness, novelty) and property prediction (membrane permeability, taste, anti-cancer activity) using MCC and AUC-ROC. Mamba generates 1-2 percent more valid and unique molecules than Mamba-2 and GPT, with fewer long-range dependency errors, while GPT yields slightly more novel structures. For property prediction, Mamba variants outperform GPT by 0.02-0.04 MCC under random splits, while scaffold splits show comparable performance. Results demonstrate that domain-specific pre-training on about 1M NPs can match models trained on datasets over 100 times larger.

We present Eureka-Audio, a compact yet high-performance audio language model that achieves competitive performance against models that are 4 to 18 times larger across a broad range of audio understanding benchmarks. Despite containing only 1.7B parameters, Eureka-Audio demonstrates strong performance on automatic speech recognition (ASR), audio understanding, and dense audio captioning, matching or surpassing multiple 7B to 30B audio and omni-modal baselines. The model adopts a unified end-to-end architecture composed of a lightweight language backbone, a Whisper-based audio encoder, and a sparsely activated Mixture-of-Experts (MoE) adapter that explicitly accounts for audio heterogeneity and alleviates cross-modal optimization conflicts under limited capacity. To further enhance paralinguistic reasoning, we introduce DataFlux, a closed loop audio instruction data synthesis and verification pipeline that constructs high quality, logically consistent supervision from raw audio. Extensive evaluations across ASR, knowledge reasoning, safety, instruction following, and paralinguistic benchmarks, demonstrate that Eureka-Audio achieves an efficient balance between computational cost and performance. These results establish Eureka Audio as a strong and practical baseline for lightweight audio understanding models.

Reinforcement learning with verifiable rewards (RLVR) has become a trending paradigm for training reasoning large language models (LLMs). However, due to the autoregressive decoding nature of LLMs, the rollout process becomes the efficiency bottleneck of RL training, consisting of up to 70\% of the total training time. In this work, we propose Quantized Reinforcement Learning (QuRL) that uses a quantized actor for accelerating the rollout. We address two challenges in QuRL. First, we propose Adaptive Clipping Range (ACR) that dynamically adjusts the clipping ratio based on the policy ratio between the full-precision actor and the quantized actor, which is essential for mitigating long-term training collapse. Second, we identify the weight update problem, where weight changes between RL steps are extremely small, making it difficult for the quantization operation to capture them effectively. We mitigate this problem through the invariant scaling technique that reduces quantization noise and increases weight update. We evaluate our method with INT8 and FP8 quantization experiments on DeepScaleR and DAPO, and achieve 20% to 80% faster rollout during training.

Reinforcement learning has become the central approach for language models (LMs) to learn from environmental reward or feedback. In practice, the environmental feedback is usually sparse and delayed. Learning from such signals is challenging, as LMs must implicitly infer how observed failures should translate into behavioral changes for future iterations. We introduce Experiential Reinforcement Learning (ERL), a training paradigm that embeds an explicit experience-reflection-consolidation loop into the reinforcement learning process. Given a task, the model generates an initial attempt, receives environmental feedback, and produces a reflection that guides a refined second attempt, whose success is reinforced and internalized into the base policy. This process converts feedback into structured behavioral revision, improving exploration and stabilizing optimization while preserving gains at deployment without additional inference cost. Across sparse-reward control environments and agentic reasoning benchmarks, ERL consistently improves learning efficiency and final performance over strong reinforcement learning baselines, achieving gains of up to +81% in complex multi-step environments and up to +11% in tool-using reasoning tasks. These results suggest that integrating explicit self-reflection into policy training provides a practical mechanism for transforming feedback into durable behavioral improvement.

In the era of large language models (LLMs), fine-tuning pretrained models has become ubiquitous. Yet the theoretical underpinning remains an open question. A central question is why only a few epochs of fine-tuning are typically sufficient to achieve strong performance on many different tasks. In this work, we approach this question by developing a statistical framework, combining rigorous early stopping theory with the attention-based Neural Tangent Kernel (NTK) for LLMs, offering new theoretical insights on fine-tuning practices. Specifically, we formally extend classical NTK theory [Jacot et al., 2018] to non-random (i.e., pretrained) initializations and provide a convergence guarantee for attention-based fine-tuning. One key insight provided by the theory is that the convergence rate with respect to sample size is closely linked to the eigenvalue decay rate of the empirical kernel matrix induced by the NTK. We also demonstrate how the framework can be used to explain task vectors for multiple tasks in LLMs. Finally, experiments with modern language models on real-world datasets provide empirical evidence supporting our theoretical insights.

Tokenization is a hardcoded compression step which remains in the training pipeline of Large Language Models (LLMs), despite a general trend towards architectures becoming increasingly end-to-end. Prior work has shown promising results at scale in bringing this compression step inside the LLMs' architecture with heuristics to draw token boundaries, and also attempts to learn these token boundaries with straight-through estimates, which treat the problem of drawing discrete token boundaries as a continuous one. We show that these token boundaries can instead be learned using score function estimates, which have tighter theoretical guarantees due to directly optimizing the problem of drawing discrete token boundaries to minimize loss. We observe that techniques from reinforcement learning, such as time discounting, are necessary to reduce the variance of this score function sufficiently to make it practicable. We demonstrate that the resultant method outperforms prior proposed straight-through estimates, both qualitatively and quantitatively at the $100$ million parameter scale.

Business environments characterized by structural demand intermittency, high variability, and multi-step planning horizons require robust and reproducible model selection mechanisms. Empirical evidence shows that no forecasting model is universally dominant and that relative rankings vary across error metrics, demand regimes, and forecast horizons, generating ambiguity in multi-SKU decision contexts. This study proposes AHSIV (Adaptive Hybrid Selector for Intermittency and Variability), a horizon-aware and regime-conditioned model selection framework designed to address horizon-induced ranking instability. The proposed approach integrates scaled and absolute error metrics adjusted through a Metric Degradation by Forecast Horizon (MDFH) procedure, structural demand classification, multi-objective Pareto dominance, and hierarchical bias refinement within a unified decision architecture. The empirical evaluation is conducted on the Walmart, M3, M4, and M5 datasets under multiple train-test partition schemes and twelve-step forecasting horizons. Results indicate that AHSIV achieves statistical equivalence with the strongest monometric baseline in terms of aggregated performance while increasing the frequency of horizon-specific best-model selection. The findings demonstrate that model selection in heterogeneous demand environments cannot be treated as a static ranking problem, and that horizon-consistent, structurally adaptive mechanisms provide a principled, operationally coherent solution for multi-SKU forecasting.

Automated Machine Learning (AutoML) has revolutionized the development of data-driven solutions; however, traditional frameworks often function as "black boxes", lacking the flexibility and transparency required for complex, real-world engineering tasks. Recent Large Language Model (LLM)-based agents have shifted toward code-driven approaches. However, they frequently suffer from hallucinated logic and logic entanglement, where monolithic code generation leads to unrecoverable runtime failures. In this paper, we present iML, a novel multi-agent framework designed to shift AutoML from black-box prompting to a code-guided, modular, and verifiable architectural paradigm. iML introduces three main ideas: (1) Code-Guided Planning, which synthesizes a strategic blueprint grounded in autonomous empirical profiling to eliminate hallucination; (2) Code-Modular Implementation, which decouples preprocessing and modeling into specialized components governed by strict interface contracts; and (3) Code-Verifiable Integration, which enforces physical feasibility through dynamic contract verification and iterative self-correction. We evaluate iML across MLE-BENCH and the newly introduced iML-BENCH, comprising a diverse range of real-world Kaggle competitions. The experimental results show iML's superiority over state-of-the-art agents, achieving a valid submission rate of 85% and a competitive medal rate of 45% on MLE-BENCH, with an average standardized performance score (APS) of 0.77. On iML-BENCH, iML significantly outperforms the other approaches by 38%-163% in APS. Furthermore, iML maintains a robust 70% success rate even under stripped task descriptions, effectively filling information gaps through empirical profiling. These results highlight iML's potential to bridge the gap between stochastic generation and reliable engineering, marking a meaningful step toward truly AutoML.

Trajectory prediction for traffic agents is critical for safe autonomous driving. However, achieving effective zero-shot generalization in previously unseen domains remains a significant challenge. Motivated by the consistent nature of kinematics across diverse domains, we aim to incorporate domain-invariant knowledge to enhance zero-shot trajectory prediction capabilities. The key challenges include: 1) effectively extracting domain-invariant scene representations, and 2) integrating invariant features with kinematic models to enable generalized predictions. To address these challenges, we propose a novel generalizable Physics-guided Causal Model (PCM), which comprises two core components: a Disentangled Scene Encoder, which adopts intervention-based disentanglement to extract domain-invariant features from scenes, and a CausalODE Decoder, which employs a causal attention mechanism to effectively integrate kinematic models with meaningful contextual information. Extensive experiments on real-world autonomous driving datasets demonstrate our method's superior zero-shot generalization performance in unseen cities, significantly outperforming competitive baselines. The source code is released at https://github.com/ZY-Zong/Physics-guided-Causal-Model.

While LLMs have seen substantial improvement in reasoning capabilities, they also sometimes overthink, generating unnecessary reasoning steps, particularly under uncertainty, given ill-posed or ambiguous queries. We introduce statistically principled early stopping methods that monitor uncertainty signals during generation to mitigate this issue. Our first approach is parametric: it models inter-arrival times of uncertainty keywords as a renewal process and applies sequential testing for stopping. Our second approach is nonparametric and provides finite-sample guarantees on the probability of halting too early on well-posed queries. We conduct empirical evaluations on reasoning tasks across several domains and models. Our results indicate that uncertainty-aware early stopping can improve both efficiency and reliability in LLM reasoning, and we observe especially significant gains for math reasoning.

Code generation has progressed more reliably than reinforcement learning, largely because code has an information structure that makes it learnable. Code provides dense, local, verifiable feedback at every token, whereas most reinforcement learning problems do not. This difference in feedback quality is not binary but graded. We propose a five-level hierarchy of learnability based on information structure and argue that the ceiling on ML progress depends less on model size than on whether a task is learnable at all. The hierarchy rests on a formal distinction among three properties of computational problems (expressibility, computability, and learnability). We establish their pairwise relationships, including where implications hold and where they fail, and present a unified template that makes the structural differences explicit. The analysis suggests why supervised learning on code scales predictably while reinforcement learning does not, and why the common assumption that scaling alone will solve remaining ML challenges warrants scrutiny.

Large language model (LLM) agents demonstrate strong performance in short-text contexts but often underperform in extended dialogues due to inefficient memory management. Existing approaches face a fundamental trade-off between efficiency and effectiveness: memory compression risks losing critical details required for complex reasoning, while retaining raw text introduces unnecessary computational overhead for simple queries. The crux lies in the limitations of monolithic memory representations and static retrieval mechanisms, which fail to emulate the flexible and proactive memory scheduling capabilities observed in humans, thus struggling to adapt to diverse problem scenarios. Inspired by the principle of cognitive economy, we propose HyMem, a hybrid memory architecture that enables dynamic on-demand scheduling through multi-granular memory representations. HyMem adopts a dual-granular storage scheme paired with a dynamic two-tier retrieval system: a lightweight module constructs summary-level context for efficient response generation, while an LLM-based deep module is selectively activated only for complex queries, augmented by a reflection mechanism for iterative reasoning refinement. Experiments show that HyMem achieves strong performance on both the LOCOMO and LongMemEval benchmarks, outperforming full-context while reducing computational cost by 92.6\%, establishing a state-of-the-art balance between efficiency and performance in long-term memory management.

Breast cancer screening programmes increasingly seek to move from one-size-fits-all interval to risk-adapted and personalized strategies. Deep learning (DL) has enabled image-based risk models with stronger 1- to 5-year prediction than traditional clinical models, but leading systems (e.g., Mirai) typically use convolutional backbones, very high-resolution inputs (>1M pixels) and simple multi-view fusion, with limited explicit modelling of contralateral asymmetry. We hypothesised that combining complementary inductive biases (convolutional and transformer-based) with explicit contralateral asymmetry modelling would allow us to match state-of-the-art 3-year risk prediction performance even when operating on substantially lower-resolution mammograms, indicating that using less detailed images in a more structured way can recover state-of-the-art accuracy. We present MamaDino, a mammography-aware multi-view attentional DINO model. MamaDino fuses frozen self-supervised DINOv3 ViT-S features with a trainable CNN encoder at 512x512 resolution, and aggregates bilateral breast information via a BilateralMixer to output a 3-year breast cancer risk score. We train on 53,883 women from OPTIMAM (UK) and evaluate on matched 3-year case-control cohorts: an in-distribution test set from four screening sites and an external out-of-distribution cohort from an unseen site. At breast-level, MamaDino matches Mirai on both internal and external tests while using ~13x fewer input pixels. Adding the BilateralMixer improves discrimination to AUC 0.736 (vs 0.713) in-distribution and 0.677 (vs 0.666) out-of-distribution, with consistent performance across age, ethnicity, scanner, tumour type and grade. These findings demonstrate that explicit contralateral modelling and complementary inductive biases enable predictions that match Mirai, despite operating on substantially lower-resolution mammograms.

We present voice2mode, a method for classification of four singing phonation modes (breathy, neutral (modal), flow, and pressed) using embeddings extracted from large self-supervised speech models. Prior work on singing phonation has relied on handcrafted signal features or task-specific neural nets; this work evaluates the transferability of speech foundation models to singing phonation classification. voice2mode extracts layer-wise representations from HuBERT and two wav2vec2 variants, applies global temporal pooling, and classifies the pooled embeddings with lightweight classifiers (SVM, XGBoost). Experiments on a publicly available soprano dataset (763 sustained vowel recordings, four labels) show that foundation-model features substantially outperform conventional spectral baselines (spectrogram, mel-spectrogram, MFCC). HuBERT embeddings obtained from early layers yield the best result (~95.7% accuracy with SVM), an absolute improvement of ~12-15% over the best traditional baseline. We also show layer-wise behaviour: lower layers, which retain acoustic/phonetic detail, are more effective than top layers specialized for Automatic Speech Recognition (ASR).

Repository-level bug localization-the task of identifying where code must be modified to fix a bug-is a critical software engineering challenge. Standard Large Language Modles (LLMs) are often unsuitable for this task due to context window limitations that prevent them from processing entire code repositories. As a result, various retrieval methods are commonly used, including keyword matching, text similarity, and simple graph-based heuristics such as Breadth-First Search. Graph Neural Networks (GNNs) offer a promising alternative due to their ability to model complex, repository-wide dependencies; however, their application has been hindered by the lack of a dedicated benchmark. To address this gap, we introduce GREPO, the first GNN benchmark for repository-scale bug localization tasks. GREPO comprises 86 Python repositories and 47294 bug-fixing tasks, providing graph-based data structures ready for direct GNN processing. Our evaluation of various GNN architectures shows outstanding performance compared to established information retrieval baselines. This work highlights the potential of GNNs for bug localization and established GREPO as a foundation resource for future research, The code is available at https://github.com/qingpingmo/GREPO.

Recent advances in agent-mediated systems have enabled a new paradigm of social network simulation, where AI agents interact with human-like autonomy. This evolution has fostered the emergence of agent-driven social networks such as Moltbook, a Reddit-like platform populated entirely by AI agents. Despite these developments, empirical comparisons between agent-driven and human-driven social networks remain scarce, limiting our understanding of how their network topologies might diverge. This paper presents the first comparative analysis of network topology on Moltbook, utilizing a comment network comprising 33,577 nodes and 697,688 edges. To provide a benchmark, we curated a parallel dataset from Reddit consisting of 7.8 million nodes and 51.8 million edges. We examine key structural differences between agent-drive and human-drive networks, specifically focusing on topological patterns and the edge formation efficacy of their respective posts. Our findings provide a foundational profile of AI-driven social structures, serving as a preliminary step toward developing more robust and authentic agent-mediated social systems.

There has been an increasing interest in topological semantics for epistemic logic, which has been shown to be useful for, e.g., modelling evidence, degrees of belief, and self-reference. We introduce a polytopological PDL capable of expressing common knowledge and various generalizations and show it has the finite model property over closure spaces but not over Cantor derivative spaces. The latter is shown by embedding a version of linear temporal logic with `past', which does not have the finite model property.

We introduce LaySPA, a reinforcement learning framework that equips large language models (LLMs) with explicit and interpretable spatial reasoning for content-aware graphic layout design. LaySPA addresses two key challenges: LLMs' limited spatial reasoning and the lack of opacity in design decision making. Instead of operating at the pixel level, we reformulate layout design as a policy learning problem over a structured textual spatial environment that explicitly encodes canvas geometry, element attributes, and inter-element relationships. LaySPA produces dual-level outputs comprising interpretable reasoning traces and structured layout specifications, enabling transparent and controllable design decision making. Layout design policy is optimized via a multi-objective spatial critique that decomposes layout quality into geometric validity, relational coherence, and aesthetic consistency, and is trained using relative group optimization to stabilize learning in open-ended design spaces. Experiments demonstrate that LaySPA improves structural validity and visual quality, outperforming larger proprietary LLMs and achieving performance comparable to specialized SOTA layout generators while requiring fewer annotated samples and reduced latency.

Algorithmic stability is a classical framework for analyzing the generalization error of learning algorithms. It predicts that an algorithm has small generalization error if it is insensitive to small perturbations in the training set such as the removal or replacement of a training point. While stability has been demonstrated for numerous well-known algorithms, this framework has had limited success in analyses of deep neural networks. In this paper we study the algorithmic stability of deep ReLU homogeneous neural networks that achieve zero training error using parameters with the smallest $L_2$ norm, also known as the minimum-norm interpolation, a phenomenon that can be observed in overparameterized models trained by gradient-based algorithms. We investigate sufficient conditions for such networks to be stable. We find that 1) such networks are stable when they contain a (possibly small) stable sub-network, followed by a layer with a low-rank weight matrix, and 2) such networks are not guaranteed to be stable even when they contain a stable sub-network, if the following layer is not low-rank. The low-rank assumption is inspired by recent empirical and theoretical results which demonstrate that training deep neural networks is biased towards low-rank weight matrices, for minimum-norm interpolation and weight-decay regularization.

Stochastic approximation (SA) is a method for finding the root of an operator perturbed by noise. There is a rich literature establishing the asymptotic normality of rescaled SA iterates under fairly mild conditions. However, these asymptotic results do not quantify the accuracy of the Gaussian approximation in finite time. In this paper, we establish explicit non-asymptotic bounds on the Wasserstein distance between the distribution of the rescaled iterate at time k and the asymptotic Gaussian limit for various choices of step-sizes including constant and polynomially decaying. As an immediate consequence, we obtain tail bounds on the error of SA iterates at any time. We obtain the sharp rates by first studying the convergence rate of the discrete Ornstein-Uhlenbeck (O-U) process driven by general noise, whose stationary distribution is identical to the limiting Gaussian distribution of the rescaled SA iterates. We believe that this is of independent interest, given its connection to sampling literature. The analysis involves adapting Stein's method for Gaussian approximation to handle the matrix weighted sum of i.i.d. random variables. The desired finite-time bounds for SA are obtained by characterizing the error dynamics between the rescaled SA iterate and the discrete time O-U process and combining it with the convergence rate of the latter process.

Recent advances in Automatic Text Recognition (ATR) have improved access to historical archives, yet a methodological divide persists between palaeographic transcriptions and normalized digital editions. While ATR models trained on more palaeographically-oriented datasets such as CATMuS have shown greater generalizability, their raw outputs remain poorly compatible with most readers and downstream NLP tools, thus creating a usability gap. On the other hand, ATR models trained to produce normalized outputs have been shown to struggle to adapt to new domains and tend to over-normalize and hallucinate. We introduce the task of Pre-Editorial Normalization (PEN), which consists in normalizing graphemic ATR output according to editorial conventions, which has the advantage of keeping an intermediate step with palaeographic fidelity while providing a normalized version for practical usability. We present a new dataset derived from the CoMMA corpus and aligned with digitized Old French and Latin editions using passim. We also produce a manually corrected gold-standard evaluation set. We benchmark this resource using ByT5-based sequence-to-sequence models on normalization and pre-annotation tasks. Our contributions include the formal definition of PEN, a 4.66M-sample silver training corpus, a 1.8k-sample gold evaluation set, and a normalization model achieving a 6.7% CER, substantially outperforming previous models for this task.

Chain-of-thought (CoT) reasoning is fundamental to modern LLM architectures and represents a critical intervention point for AI safety. However, CoT reasoning may exhibit failure modes that we note as pathologies, which prevent it from being useful for monitoring. Prior work has identified three distinct pathologies: post-hoc rationalization, where models generate plausible explanations backwards from predetermined answers; encoded reasoning, where intermediate steps conceal information within seemingly interpretable text; and internalized reasoning, where models replace explicit reasoning with meaningless filler tokens while computing internally. To better understand and discriminate between these pathologies, we create a set of concrete metrics that are simple to implement, computationally inexpensive, and task-agnostic. To validate our approach, we develop model organisms deliberately trained to exhibit specific CoT pathologies. Our work provides a practical toolkit for assessing CoT pathologies, with direct implications for training-time monitoring.

In this paper, we propose RPGD (RANSAC-P3P Gradient Descent), a human-pose-driven extrinsic calibration framework that robustly aligns MoCap-based 3D skeletal data with monocular or multi-view RGB cameras using only natural human motion. RPGD formulates extrinsic calibration as a coarse-to-fine problem tailored to human poses, combining the global robustness of RANSAC-P3P with Gradient-Descent-based refinement. We evaluate RPGD on three large-scale public 3D HPE datasets as well as on a self-collected in-the-wild dataset. Experimental results demonstrate that RPGD consistently recovers extrinsic parameters with accuracy comparable to the provided ground truth, achieving sub-pixel MPJPE reprojection error even in challenging, noisy settings. These results indicate that RPGD provides a practical and automatic solution for reliable extrinsic calibration of large-scale 3D HPE dataset collection.

Recent advances in speech language models, such as GPT-4o Voice Mode and Gemini Live, have demonstrated promising speech generation capabilities. Nevertheless, the aesthetic naturalness of the synthesized audio still lags behind that of human speech. Enhancing generation quality requires a reliable evaluator of speech naturalness. However, existing naturalness evaluators typically regress raw audio to scalar scores, offering limited interpretability of the evaluation and moreover fail to generalize to speech across different taxonomies. Inspired by recent advances in generative reward modeling, we propose the Generative Speech Reward Model (GSRM), a reasoning-centric reward model tailored for speech. The GSRM is trained to decompose speech naturalness evaluation into an interpretable acoustic feature extraction stage followed by feature-grounded chain-of-thought reasoning, enabling explainable judgments. To achieve this, we curated a large-scale human feedback dataset comprising 31k expert ratings and an out-of-domain benchmark of real-world user-assistant speech interactions. Experiments show that GSRM substantially outperforms existing speech naturalness predictors, achieving model-human correlation of naturalness score prediction that approaches human inter-rater consistency. We further show how GSRM can improve the naturalness of speech LLM generations by serving as an effective verifier for online RLHF.

Retrieval Augmented Generation (RAG) is a powerful approach for enhancing the factual grounding of language models by integrating external knowledge. While widely studied for large language models, the optimization of RAG for Small Language Models (SLMs) remains a critical research gap, particularly in complex, multi-hop question-answering tasks that require sophisticated reasoning. In these systems, prompt template design is a crucial yet under-explored factor influencing performance. This paper presents a large-scale empirical study to investigate this factor, evaluating 24 different prompt templates on the HotpotQA dataset. The set includes a standard RAG prompt, nine well-formed techniques from the literature, and 14 novel hybrid variants, all tested on two prominent SLMs: Qwen2.5-3B Instruct and Gemma3-4B-It. Our findings, based on a test set of 18720 instances, reveal significant performance gains of up to 83% on Qwen2.5 and 84.5% on Gemma3-4B-It, yielding an improvement of up to 6% for both models compared to the Standard RAG prompt. This research also offers concrete analysis and actionable recommendations for designing effective and efficient prompts for SLM-based RAG systems, practically for deployment in resource-constrained environments.

We present a font classification system capable of identifying 394 font families from rendered text images. Our approach fine-tunes a DINOv2 Vision Transformer using Low-Rank Adaptation (LoRA), achieving approximately 86% top-1 accuracy while training fewer than 1% of the model's 87.2M parameters. We introduce a synthetic dataset generation pipeline that renders Google Fonts at scale with diverse augmentations including randomized colors, alignment, line wrapping, and Gaussian noise, producing training images that generalize to real-world typographic samples. The model incorporates built-in preprocessing to ensure consistency between training and inference, and is deployed as a HuggingFace Inference Endpoint. We release the model, dataset, and full training pipeline as open-source resources.

Graph property detection aims to determine whether a graph exhibits certain structural properties, such as being Hamiltonian. Recently, learning-based approaches have shown great promise by leveraging data-driven models to detect graph properties efficiently. In particular, vision-based methods offer a visually intuitive solution by processing the visualizations of graphs. However, existing vision-based methods rely on fixed visual graph layouts, and therefore, the expressiveness of their pipeline is restricted. To overcome this limitation, we propose VSAL, a vision-based framework that incorporates an adaptive layout generator capable of dynamically producing informative graph visualizations tailored to individual instances, thereby improving graph property detection. Extensive experiments demonstrate that VSAL outperforms state-of-the-art vision-based methods on various tasks such as Hamiltonian cycle, planarity, claw-freeness, and tree detection.

Multi-agent systems are designed to deal with open, distributed systems with unpredictable dynamics, which makes them inherently hard to test. The value of using simulation for this purpose is recognized in the literature, although achieving sufficient fidelity (i.e., the degree of similarity between the simulation and the real-world system) remains a challenging task. This is exacerbated when dealing with cognitive agent models, such as the Belief Desire Intention (BDI) model, where the agent codebase is not suitable to run unchanged in simulation environments, thus increasing the reality gap between the deployed and simulated systems. We argue that BDI developers should be able to test in simulation the same specification that will be later deployed, with no surrogate representations. Thus, in this paper, we discuss how the control flow of BDI agents can be mapped onto a Discrete Event Simulation (DES), showing that such integration is possible at different degrees of granularity. We substantiate our claims by producing an open-source prototype integration between two pre-existing tools (JaKtA and Alchemist), showing that it is possible to produce a simulation-based testing environment for distributed BDI} agents, and that different granularities in mapping BDI agents over DESs may lead to different degrees of fidelity.

In many scientific settings, acquiring complete observations of PDE coefficients and solutions can be expensive, hazardous, or impossible. Recent diffusion-based methods can reconstruct fields given partial observations, but require complete observations for training. We introduce Ambient Physics, a framework for learning the joint distribution of coefficient-solution pairs directly from partial observations, without requiring a single complete observation. The key idea is to randomly mask a subset of already-observed measurements and supervise on them, so the model cannot distinguish "truly unobserved" from "artificially unobserved", and must produce plausible predictions everywhere. Ambient Physics achieves state-of-the-art reconstruction performance. Compared with prior diffusion-based methods, it achieves a 62.51$\%$ reduction in average overall error while using 125$\times$ fewer function evaluations. We also identify a "one-point transition": masking a single already-observed point enables learning from partial observations across architectures and measurement patterns. Ambient Physics thus enables scientific progress in settings where complete observations are unavailable.

We formulate Ensemble-Conditional Gaussian Processes (Ens-CGP), a finite-dimensional synthesis that centers ensemble-based inference on the conditional Gaussian law. Conditional Gaussian processes (CGP) arise directly from Gaussian processes under conditioning and, in linear-Gaussian settings, define the full posterior distribution for a Gaussian prior and linear observations. Classical Kalman filtering is a recursive algorithm that computes this same conditional law under dynamical assumptions; the conditional Gaussian law itself is therefore the underlying representational object, while the filter is one computational realization. In this sense, CGP provides the probabilistic foundation for Kalman-type methods as well as equivalent formulations as a strictly convex quadratic program (MAP estimation), RKHS-regularized regression, and classical regularization. Ens-CGP is the ensemble instantiation of this object, obtained by treating empirical ensemble moments as a (possibly low-rank) Gaussian prior and performing exact conditioning. By separating representation (GP -> CGP -> Ens-CGP) from computation (Kalman filters, EnKF variants, and iterative ensemble schemes), the framework links an earlier-established representational foundation for inference to ensemble-derived priors and clarifies the relationships among probabilistic, variational, and ensemble perspectives.

The growing importance of culturally-aware natural language processing systems has led to an increasing demand for resources that capture sociopragmatic phenomena across diverse languages. Nevertheless, Arabic-language resources for politeness detection remain under-explored, despite the rich and complex politeness expressions embedded in Arabic communication. In this paper, we introduce ADAB (Arabic Politeness Dataset), a new annotated Arabic dataset collected from four online platforms, including social media, e-commerce, and customer service domains, covering Modern Standard Arabic and multiple dialects (Gulf, Egyptian, Levantine, and Maghrebi). The dataset was annotated based on Arabic linguistic traditions and pragmatic theory, resulting in three classes: polite, impolite, and neutral. It contains 10,000 samples with linguistic feature annotations across 16 politeness categories and achieves substantial inter-annotator agreement (kappa = 0.703). We benchmark 40 model configurations, including traditional machine learning, transformer-based models, and large language models. The dataset aims to support research on politeness-aware Arabic NLP.

Large language models (LLMs) are being deployed across the Global South, where everyday use involves low-resource languages, code-mixing, and culturally specific norms. Yet safety pipelines, benchmarks, and alignment still largely target English and a handful of high-resource languages, implicitly assuming safety and factuality ''transfer'' across languages. Evidence increasingly shows they do not. We synthesize recent findings indicating that (i) safety guardrails weaken sharply on low-resource and code-mixed inputs, (ii) culturally harmful behavior can persist even when standard toxicity scores look acceptable, and (iii) English-only knowledge edits and safety patches often fail to carry over to low-resource languages. In response, we outline a practical agenda for researchers and students in the Global South: parameter-efficient safety steering, culturally grounded evaluation and preference data, and participatory workflows that empower local communities to define and mitigate harm. Our aim is to make multilingual safety a core requirement-not an add-on-for equitable AI in underrepresented regions.

Deploying heterogeneous robot teams to accomplish multiple tasks over extended time horizons presents significant computational challenges for task allocation and planning. In this paper, we present a comprehensive, time-extended, offline heterogeneous multi-robot task allocation framework, TRAITS, which we believe to be the first that can cope with the provisioning of exhaustible traits under battery and temporal constraints. Specifically, we introduce a nonlinear programming-based trait distribution module that can optimize the trait-provisioning rate of coalitions to yield feasible and time-efficient solutions. TRAITS provides a more accurate feasibility assessment and estimation of task execution times and makespan by leveraging trait-provisioning rates while optimizing battery consumption -- an advantage that state-of-the-art frameworks lack. We evaluate TRAITS against two state-of-the-art frameworks, with results demonstrating its advantage in satisfying complex trait and battery requirements while remaining computationally tractable.

Hierarchical Reinforcement Learning (HRL) frameworks like Option-Critic (OC) and Multi-updates Option Critic (MOC) have introduced significant advancements in learning reusable options. However, these methods underperform in multi-goal environments with sparse rewards, where actions must be linked to temporally distant outcomes. To address this limitation, we first propose MOC-HER, which integrates the Hindsight Experience Replay (HER) mechanism into the MOC framework. By relabeling goals from achieved outcomes, MOC-HER can solve sparse reward environments that are intractable for the original MOC. However, this approach is insufficient for object manipulation tasks, where the reward depends on the object reaching the goal rather than on the agent's direct interaction. This makes it extremely difficult for HRL agents to discover how to interact with these objects. To overcome this issue, we introduce Dual Objectives Hindsight Experience Replay (2HER), a novel extension that creates two sets of virtual goals. In addition to relabeling goals based on the object's final state (standard HER), 2HER also generates goals from the agent's effector positions, rewarding the agent for both interacting with the object and completing the task. Experimental results in robotic manipulation environments show that MOC-2HER achieves success rates of up to 90%, compared to less than 11% for both MOC and MOC-HER. These results highlight the effectiveness of our dual objective relabeling strategy in sparse reward, multi-goal tasks.

Learning in the presence of missing data can result in biased predictions and poor generalizability, among other difficulties, which data imputation methods only partially address. In neural networks, activation functions significantly affect performance yet typical options (e.g., ReLU, Swish) operate only on feature values and do not account for missingness indicators or confidence scores. We propose Three-Channel Evolved Activations (3C-EA), which we evolve using Genetic Programming to produce multivariate activation functions f(x, m, c) in the form of trees that take (i) the feature value x, (ii) a missingness indicator m, and (iii) an imputation confidence score c. To make these activations useful beyond the input layer, we introduce ChannelProp, an algorithm that deterministically propagates missingness and confidence values via linear layers based on weight magnitudes, retaining reliability signals throughout the network. We evaluate 3C-EA and ChannelProp on datasets with natural and injected (MCAR/MAR/MNAR) missingness at multiple rates under identical preprocessing and splits. Results indicate that integrating missingness and confidence inputs into the activation search improves classification performance under missingness.

This paper presents the history of the online simulation program Java-DSP (J-DSP) and the most recent function development and deployment. J-DSP was created to support online laboratories in DSP classes and was first deployed in our ASU DSP class in 2000. The development of the program and its extensions was supported by several NSF grants including CCLI and IUSE. The web-based software was developed by our team in Java and later transitioned to the more secure HTML5 environment. J-DSP supports laboratory exercises on: digital filters and their design, the FFT and its utility in spectral analysis, machine learning for signal classification, and more recently online simulations with the Quantum Fourier Transform. Throughout the J-DSP development and deployment of this tool and its associated laboratory exercises, we documented evaluations. Mobile versions of the program for iOS and Android were also developed. J-DSP is used to this day in several universities, and specific functions of the program have been used in NSF REU, IRES and RET workforce development and high school outreach.

The overarching research direction of this work is the development of a ''Responsible Intelligence'' framework designed to reconcile the immense generative power of Large Language Models (LLMs) with the stringent requirements of real-world deployment. As these models become a transformative force in artificial intelligence, there is an urgent need to move beyond general-purpose architectures toward systems that are contextually aware, inherently safer, and deeply respectful of global cultural nuances. This research navigates three interconnected threads: domain adaptation to ensure technical precision, ethical rigor to mitigate adversarial vulnerabilities, and cultural/multilingual alignment to promote global inclusivity. The methodological trajectory moves from classical supervised adaptation for task-specific demands to decoding-time alignment for safety, finally leveraging human feedback and preference modeling to achieve sociolinguistic acuity.

Tasks ranging from sleep staging to clinical diagnosis traditionally rely on standard polysomnography (PSG) devices, bedside monitors and wearable devices, which capture diverse nocturnal biosignals (e.g., EEG, EOG, ECG, SpO$_2$). However, heterogeneity across devices and frequent sensor dropout pose significant challenges for unified modelling of these multimodal signals. We present \texttt{sleep2vec}, a foundation model for diverse and incomplete nocturnal biosignals that learns a shared representation via cross-modal alignment. \texttt{sleep2vec} is contrastively pre-trained on 42,249 overnight recordings spanning nine modalities using a \textit{Demography, Age, Site \& History-aware InfoNCE} objective that incorporates physiological and acquisition metadata (\textit{e.g.}, age, gender, recording site) to dynamically weight negatives and mitigate cohort-specific shortcuts. On downstream sleep staging and clinical outcome assessment, \texttt{sleep2vec} consistently outperforms strong baselines and remains robust to any subset of available modalities and sensor dropout. We further characterize, to our knowledge for the first time, scaling laws for nocturnal biosignals with respect to modality diversity and model capacity. Together, these results show that unified cross-modal alignment, coupled with principled scaling, enables label-efficient, general-purpose modelling of real-world nocturnal biosignals.

A deep research agent produces a fluent scientific report in minutes; a careful reader then tries to verify the main claims and discovers the real cost is not reading, but tracing: which sentence is supported by which passage, what was ignored, and where evidence conflicts. We argue that as research generation becomes cheap, auditability becomes the bottleneck, and the dominant risk shifts from isolated factual errors to scientifically styled outputs whose claim-evidence links are weak, missing, or misleading. This perspective proposes claim-level auditability as a first-class design and evaluation target for deep research agents, distills recurring long-horizon failure modes (objective drift, transient constraints, and unverifiable inference), and introduces the Auditable Autonomous Research (AAR) standard, a compact measurement framework that makes auditability testable via provenance coverage, provenance soundness, contradiction transparency, and audit effort. We then argue for semantic provenance with protocolized validation: persistent, queryable provenance graphs that encode claim--evidence relations (including conflicts) and integrate continuous validation during synthesis rather than after publication, with practical instrumentation patterns to support deployment at scale.

Modern online experimentation faces two bottlenecks: scarce traffic forces tough choices on which variants to test, and post-hoc insight extraction is manual, inconsistent, and often content-agnostic. Meanwhile, organizations underuse historical A/B results and rich content embeddings that could guide prioritization and creative iteration. We present a unified framework to (i) prioritize which variants to test, (ii) explain why winners win, and (iii) surface targeted opportunities for new, higher-potential variants. Leveraging treatment embeddings and historical outcomes, we train a CTR ranking model with fixed effects for contextual shifts that scores candidates while balancing value and content diversity. For better interpretability and understanding, we project treatments onto curated semantic marketing attributes and re-express the ranker in this space via a sign-consistent, sparse constrained Lasso, yielding per-attribute coefficients and signed contributions for visual explanations, top-k drivers, and natural-language insights. We then compute an opportunity index combining attribute importance (from the ranker) with under-expression in the current experiment to flag missing, high-impact attributes. Finally, LLMs translate ranked opportunities into concrete creative suggestions and estimate both learning and conversion potential, enabling faster, more informative, and more efficient test cycles. These components have been built into a real Adobe product, called \textit{Experimentation Accelerator}, to provide AI-based insights and opportunities to scale experimentation for customers. We provide an evaluation of the performance of the proposed framework on some real-world experiments by Adobe business customers that validate the high quality of the generation pipeline.

Formal specifications are crucial for building verifiable and dependable software systems, yet generating accurate and verifiable specifications for real-world C programs remains challenging. This paper empirically evaluates the extent to which formal-analysis tools can automatically generate and verify ACSL specifications without human or learning-based assistance. We conduct a controlled study on a recently released dataset of 506 C programs, repurposing it from interactive, developer-driven workflows to an automated evaluation setting. Five ACSL generation systems are compared: a rule-based Python script, Frama-C's RTE plugin, and three large language models--DeepSeek-V3.2, GPT-5.2, and OLMo 3.1 32B Instruct. All generated specifications are verified under identical conditions using the Frama-C WP plugin powered by multiple SMT solvers, allowing a direct comparison of annotation quality, solver sensitivity, and proof stability. Our results provide new empirical evidence on the capabilities and limitations of automated ACSL generation, complementing prior survey-based work.

We propose a sequential test for detecting arbitrary distribution shifts that allows conformal test martingales (CTMs) to work under a fixed, reference-conditional setting. Existing CTM detectors construct test martingales by continually growing a reference set with each incoming sample, using it to assess how atypical the new sample is relative to past observations. While this design yields anytime-valid type-I error control, it suffers from test-time contamination: after a change, post-shift observations enter the reference set and dilute the evidence for distribution shift, increasing detection delay and reducing power. In contrast, our method avoids contamination by design by comparing each new sample to a fixed null reference dataset. Our main technical contribution is a robust martingale construction that remains valid conditional on the null reference data, achieved by explicitly accounting for the estimation error in the reference distribution induced by the finite reference set. This yields anytime-valid type-I error control together with guarantees of asymptotic power one and bounded expected detection delay. Empirically, our method detects shifts faster than standard CTMs, providing a powerful and reliable distribution-shift detector.

We introduce a flexible framework based on response theory and score matching to suppress spurious, noncausal dependencies in reduced-order neural emulators of turbulent systems, focusing on climate dynamics as a proof-of-concept. We showcase the approach using the stochastic Charney-DeVore model as a relevant prototype for low-frequency atmospheric variability. We show that the resulting causal constraints enhance neural emulators' ability to respond to both weak and strong external forcings, despite being trained exclusively on unforced data. The approach is broadly applicable to modeling complex turbulent dynamical systems in reduced spaces and can be readily integrated into general neural network architectures.

[Context] Online Recruitment and Selection (R&S) processes are often the first point of contact between early-career software engineers and the tech industry. Yet many candidates experience these processes as opaque, inefficient, or even discouraging. While prior research has extensively documented the flaws and biases in online tech hiring, little is known about the practices that create positive candidate experiences. [Objective & Method] This paper explores such practices, referred to as Constructive Patterns (CPs), from the perspective of early-career software engineers. Guided by Applicant Attribution-Reaction Theory, we conducted 22 semi-structured interviews in which participants collectively described over 470 online R&S experiences. [Results] Through thematic analysis, we identified 22 CPs that reflect positive practices such as comprehensive and transparent job advertisements (CP01), specific and developmental feedback (CP03), humanized and respectful interaction (CP06), and framing the process as a two-way street (CP18). [Conclusion] Our findings extend the conversation on tech hiring beyond diagnosing dysfunctions toward designing for human-centered and growth-oriented candidate experiences. The resulting catalog of CPs provides a concrete and empirically grounded resource for organizations seeking to attract and support early-career software engineers more effectively.

Severe aortic stenosis is a common and life-threatening condition in elderly patients, often treated with Transcatheter Aortic Valve Implantation (TAVI). Despite procedural advances, paravalvular aortic regurgitation (PVR) remains one of the most frequent post-TAVI complications, with a proven impact on long-term prognosis. In this work, we investigate the potential of deep learning to predict the occurrence of PVR from preoperative cardiac CT. To this end, a dataset of preoperative TAVI patients was collected, and 3D convolutional neural networks were trained on isotropic CT volumes. The results achieved suggest that volumetric deep learning can capture subtle anatomical features from pre-TAVI imaging, opening new perspectives for personalized risk assessment and procedural optimization. Source code is available at https://github.com/EIDOSLAB/tavi.

Large language model (LLM) agents are increasingly deployed in personalized tasks involving sensitive, context-dependent information, where privacy violations may arise in agents' action due to the implicitness of contextual privacy. Existing approaches rely on external, inference-time interventions which are brittle, scenario-specific, and may expand the privacy attack surface. We propose PrivAct, a contextual privacy-aware multi-agent learning framework that internalizes contextual privacy preservation directly into models' generation behavior for privacy-compliant agentic actions. By embedding privacy preferences into each agent, PrivAct enhances system-wide contextual integrity while achieving a more favorable privacy-helpfulness tradeoff. Experiments across multiple LLM backbones and benchmarks demonstrate consistent improvements in contextual privacy preservation, reducing leakage rates by up to 12.32% while maintaining comparable helpfulness, as well as zero-shot generalization and robustness across diverse multi-agent topologies. Code is available at https://github.com/chengyh23/PrivAct.

Speculative decoding has rapidly emerged as a leading approach for accelerating language model (LM) inference, as it offers substantial speedups while yielding identical outputs. This relies upon a small draft model, tasked with predicting the outputs of the target model. State-of-the-art speculative decoding methods use a draft model consisting of a single decoder layer and output embedding matrix, with the latter dominating drafting time for the latest LMs. Recent work has sought to address this output distribution bottleneck by reducing the vocabulary of the draft model. Although this can improve throughput, it compromises speculation effectiveness when the target token is out-of-vocabulary. In this paper, we argue for vocabulary speculation as an alternative to a reduced vocabulary. We propose SpecVocab, an efficient and effective method that selects a vocabulary subset per decoding step. Across a variety of tasks, we demonstrate that SpecVocab can achieve a higher acceptance length than state-of-the-art speculative decoding approach, EAGLE-3. Notably, this yields up to an 8.1% increase in average throughput over EAGLE-3.

Large Language Models (LLMs) have developed rapidly and are widely applied to both general-purpose and professional tasks to assist human users. However, they still struggle to comprehend and respond to the true user needs when intentions and instructions are imprecisely conveyed, leading to a divergence between subjective user believes and true environment states. Resolving this epistemic divergence requires Theory of Mind (ToM), yet existing ToM evaluations for LLMs primarily focus on isolated belief inference, overlooking its functional utility in real-world interaction. To this end, we formalize ToM for LLMs as a mechanism for epistemic divergence detection and resolution, and propose a benchmark, \benchname, to assess how models reconcile user beliefs and profiles in practice. Results across 11 leading models reveal a significant limitation to identify underlying cognitive gaps that impede task success. To bridge this gap, we further curate a trajectory-based ToM dataset linking belief tracking with task-related state inference. The model trained on this data via reinforcement learning shows consistent improvement in reasoning about user mental states, leading to enhanced downstream performance. Our work highlights the practical value of ToM as an essential interaction-level mechanism rather than as a standalone reasoning skill.

Recent advances in auto-regressive transformers have achieved remarkable success in generative modeling. However, text-to-3D generation remains challenging, primarily due to bottlenecks in learning discrete 3D representations. Specifically, existing approaches often suffer from information loss during encoding, causing representational distortion before the quantization process. This effect is further amplified by vector quantization, ultimately degrading the geometric coherence of text-conditioned 3D shapes. Moreover, the conventional two-stage training paradigm induces an objective mismatch between reconstruction and text-conditioned auto-regressive generation. To address these issues, we propose View-aware Auto-Regressive 3D (VAR-3D), which intergrates a view-aware 3D Vector Quantized-Variational AutoEncoder (VQ-VAE) to convert the complex geometric structure of 3D models into discrete tokens. Additionally, we introduce a rendering-supervised training strategy that couples discrete token prediction with visual reconstruction, encouraging the generative process to better preserve visual fidelity and structural consistency relative to the input text. Experiments demonstrate that VAR-3D significantly outperforms existing methods in both generation quality and text-3D alignment.

AI is reshaping academic research, yet its role in peer review remains polarising and contentious. Advocates see its potential to reduce reviewer burden and improve quality, while critics warn of risks to fairness, accountability, and trust. At ICLR 2025, an official AI feedback tool was deployed to provide reviewers with post-review suggestions. We studied this deployment through surveys and interviews, investigating how reviewers engaged with the tool and perceived its usability and impact. Our findings surface both opportunities and tensions when AI augments in peer review. This work contributes the first empirical evidence of such an AI tool in a live review process, documenting how reviewers respond to AI-generated feedback in a high-stakes review context. We further offer design implications for AI-assisted reviewing that aim to enhance quality while safeguarding human expertise, agency, and responsibility.

This study examines the acquisition of English irregular inflections by Yemeni learners of English as a second language (L2), utilizing a Universal Grammar (UG) approach. Within the UG approach, the study considers Feature Reassembly Hypothesis (FRH) (Lardiere, 2008, 2009) part of UG, focusing on the roles of first language (L1) transfer and L2 developmental influence. It analyzes learner errors across two developmental stages. Stage 1 data reveal a dominant influence of L1 transfer, particularly in phonological and structural mismatches, while stage 2 data demonstrate increased learner sensitivity to UG properties and morphological reconfiguration toward the target language. Findings reveal that errors in irregular inflectional morphology are attributed to both interlingual and intralingual sources, with overgeneralization of L2 rules as a common developmental strategy. Statistical analysis, including a one-way ANOVA, indicates significant improvement in the production of well-formed irregular inflections from stage 1 to stage 2, underscoring learners' continued access to UG. However, persistent difficulties with consonant change, zero-morpheme, and -a plural inflections suggest that limited exposure, ineffective input modeling, and insufficient instructional quality constrain full UG access. The study concludes that while L1 transfer and L2 developmental factors influence initial stages of acquisition, appropriate linguistic input and instruction are critical for facilitating UG-driven feature reassembly in adult L2 learners.

We introduce Pawsterior, a variational flow-matching framework for improved and extended simulation-based inference (SBI). Many SBI problems involve posteriors constrained by structured domains, such as bounded physical parameters or hybrid discrete-continuous variables, yet standard flow-matching methods typically operate in unconstrained spaces. This mismatch leads to inefficient learning and difficulty respecting physical constraints. Our contributions are twofold. First, generalizing the geometric inductive bias of CatFlow, we formalize endpoint-induced affine geometric confinement, a principle that incorporates domain geometry directly into the inference process via a two-sided variational model. This formulation improves numerical stability during sampling and leads to consistently better posterior fidelity, as demonstrated by improved classifier two-sample test performance across standard SBI benchmarks. Second, and more importantly, our variational parameterization enables SBI tasks involving discrete latent structure (e.g., switching systems) that are fundamentally incompatible with conventional flow-matching approaches. By addressing both geometric constraints and discrete latent structure, Pawsterior extends flow-matching to a broader class of structured SBI problems that were previously inaccessible.

Document-to-table (Doc2Table) extraction derives structured tables from unstructured documents under a target schema, enabling reliable and verifiable SQL-based data analytics. Although large language models (LLMs) have shown promise in flexible information extraction, their ability to produce precisely structured tables remains insufficiently understood, particularly for indirect extraction that requires complex capabilities such as reasoning and conflict resolution. Existing benchmarks neither explicitly distinguish nor comprehensively cover the diverse capabilities required in Doc2Table extraction.We argue that a capability-aware benchmark is essential for systematic evaluation. However, constructing such benchmarks using human-annotated document-table pairs is costly, difficult to scale, and limited in capability coverage. To address this, we adopt a reverse Table2Doc paradigm and design a multi-agent synthesis workflow to generate documents from ground-truth tables. Based on this approach, we present DTBench, a synthetic benchmark that adopts a proposed two-level taxonomy of Doc2Table capabilities, covering 5 major categories and 13 subcategories. We evaluate several mainstream LLMs on DTBench, and demonstrate substantial performance gaps across models, as well as persistent challenges in reasoning, faithfulness, and conflict resolution. DTBench provides a comprehensive testbed for data generation and evaluation, facilitating future research on Doc2Table extraction. The benchmark is publicly available at https://github.com/ZJU-DAILY/DTBench.

Physics-Informed Neural Networks present a novel approach in SciML that integrates physical laws in the form of partial differential equations directly into the NN through soft constraints in the loss function. This work studies the application of PINNs to solve a one dimensional coupled electro-elastodynamic system modeling linear piezoelectricity in stress-charge form, governed by elastodynamic and electrodynamic equations. Our simulation employs a feedforward architecture, mapping space-time coordinates to mechanical displacement and electric potential. Our PINN model achieved global relative L2 errors of 2.34 and 4.87 percent for displacement and electric potential respectively. The results validate PINNs as effective mesh free solvers for coupled time-dependent PDE systems, though challenges remain regarding error accumulation and stiffness in coupled eigenvalue systems.

Learning expressive and efficient policy functions is a promising direction in reinforcement learning (RL). While flow-based policies have recently proven effective in modeling complex action distributions with a fast deterministic sampling process, they still face a trade-off between expressiveness and computational burden, which is typically controlled by the number of flow steps. In this work, we propose mean velocity policy (MVP), a new generative policy function that models the mean velocity field to achieve the fastest one-step action generation. To ensure its high expressiveness, an instantaneous velocity constraint (IVC) is introduced on the mean velocity field during training. We theoretically prove that this design explicitly serves as a crucial boundary condition, thereby improving learning accuracy and enhancing policy expressiveness. Empirically, our MVP achieves state-of-the-art success rates across several challenging robotic manipulation tasks from Robomimic and OGBench. It also delivers substantial improvements in training and inference speed over existing flow-based policy baselines.

We present an end-to-end framework for systematic evaluation of LLM-generated smart contracts from natural-language specifications. The system parses contractual text into structured schemas, generates Solidity code, and performs automated quality assessment through compilation and security checks. Using CrewAI-style agent teams with iterative refinement, the pipeline produces structured artifacts with full provenance metadata. Quality is measured across five dimensions, including functional completeness, variable fidelity, state-machine correctness, business-logic fidelity, and code quality aggregated into composite scores. The framework supports paired evaluation against ground-truth implementations, quantifying alignment and identifying systematic error modes such as logic omissions and state transition inconsistencies. This provides a reproducible benchmark for empirical research on smart contract synthesis quality and supports extensions to formal verification and compliance checking.

Time series anomaly detection is critical in many real-world applications, where effective solutions must localize anomalous regions and support reliable decision-making under complex settings. However, most existing methods frame anomaly detection as a purely discriminative prediction task with fixed feature inputs, rather than an evidence-driven diagnostic process. As a result, they often struggle when anomalies exhibit strong context dependence or diverse patterns. We argue that these limitations stem from the lack of adaptive feature preparation, reasoning-aware detection, and iterative refinement during inference. To address these challenges, we propose AnomaMind, an agentic time series anomaly detection framework that reformulates anomaly detection as a sequential decision-making process. AnomaMind operates through a structured workflow that progressively localizes anomalous intervals in a coarse-to-fine manner, augments detection through multi-turn tool interactions for adaptive feature preparation, and refines anomaly decisions via self-reflection. The workflow is supported by a set of reusable tool engines, enabling context-aware diagnostic analysis. A key design of AnomaMind is an explicitly designed hybrid inference mechanism for tool-augmented anomaly detection. In this mechanism, general-purpose models are responsible for autonomous tool interaction and self-reflective refinement, while core anomaly detection decisions are learned through reinforcement learning under verifiable workflow-level feedback, enabling task-specific optimization within a flexible reasoning framework. Extensive experiments across diverse settings demonstrate that AnomaMind consistently improves anomaly detection performance. The code is available at https://anonymous.4open.science/r/AnomaMind.

Untrained neural networks (UNNs) offer high-fidelity electromagnetic inverse scattering reconstruction but are computationally limited by high-dimensional spatial-domain optimization. We propose a Real-Time Physics-Driven Fourier-Spectral (PDF) solver that achieves sub-second reconstruction through spectral-domain dimensionality reduction. By expanding induced currents using a truncated Fourier basis, the optimization is confined to a compact low-frequency parameter space supported by scattering measurements. The solver integrates a contraction integral equation (CIE) to mitigate high-contrast nonlinearity and a contrast-compensated operator (CCO) to correct spectral-induced attenuation. Furthermore, a bridge-suppressing loss is formulated to enhance boundary sharpness between adjacent scatterers. Numerical and experimental results demonstrate a 100-fold speedup over state-of-the-art UNNs with robust performance under noise and antenna uncertainties, enabling real-time microwave imaging applications.

Large language models spend most of their inference cost on attention over long contexts, yet empirical behavior suggests that only a small subset of tokens meaningfully contributes to each query. We formalize this phenomenon by modeling attention as a projection onto the convex hull of key vectors and analyzing its entropic (softmax-like) relaxation. Our main theoretical contribution is a face-stability theorem showing that, under a strict complementarity margin (a support gap (Δ) certified by KKT multipliers), entropic attention concentrates on a constant-size active face: the total mass assigned to inactive tokens decays exponentially as (\exp(-Ω(Δ/\varepsilon))), while the error on the active face scales linearly in the temperature/regularization parameter (\varepsilon). This yields a practical criterion for when sparse long-context decoding is safe and provides a principled knob to trade accuracy for compute. Building on these guarantees, we introduce Vashista Sparse Attention, a drop-in mechanism that maintains a small candidate set per query through a paging-style context selection strategy compatible with modern inference stacks. Across long-context evaluations, we observe stable constant-size effective support, strong wall-clock speedups, and minimal quality degradation in the regimes predicted by the support-gap diagnostics. Finally, we discuss deployment implications for privacy-sensitive and air-gapped settings, where interchangeable attention modules enable predictable latency and cost without external retrieval dependencies.

Time series forecasting has long been dominated by model-centric approaches that formulate prediction as a single-pass mapping from historical observations to future values. Despite recent progress, such formulations often struggle in complex and evolving settings, largely because most forecasting models lack the ability to autonomously acquire informative evidence, reason about potential future changes, or revise predictions through iterative decision processes. In this work, we propose Cast-R1, a learned time series forecasting framework that reformulates forecasting as a sequential decision-making problem. Cast-R1 introduces a memory-based state management mechanism that maintains decision-relevant information across interaction steps, enabling the accumulation of contextual evidence to support long-horizon reasoning. Building on this formulation, forecasting is carried out through a tool-augmented agentic workflow, in which the agent autonomously interacts with a modular toolkit to extract statistical features, invoke lightweight forecasting models for decision support, perform reasoning-based prediction, and iteratively refine forecasts through self-reflection. To train Cast-R1, we adopt a two-stage learning strategy that combines supervised fine-tuning with multi-turn reinforcement learning, together with a curriculum learning scheme that progressively increases task difficulty to improve policy learning. Extensive experiments on multiple real-world time series datasets demonstrate the effectiveness of Cast-R1. We hope this work provides a practical step towards further exploration of agentic paradigms for time series modeling. Our code is available at https://github.com/Xiaoyu-Tao/Cast-R1-TS.

Ovarian tumour management has increasingly relied on multidisciplinary tumour board (MDT) deliberation to address treatment complexity and disease heterogeneity. However, most patients worldwide lack access to timely expert consensus, particularly in resource-constrained centres where MDT resources are scarce or unavailable. Here we present OMGs (Ovarian tumour Multidisciplinary intelligent aGent System), a multi-agent AI framework where domain-specific agents deliberate collaboratively to integrate multidisciplinary evidence and generate MDT-style recommendations with transparent rationales. To systematically evaluate MDT recommendation quality, we developed SPEAR (Safety, Personalization, Evidence, Actionability, Robustness) and validated OMGs across diverse clinical scenarios spanning the care continuum. In multicentre re-evaluation, OMGs achieved performance comparable to expert MDT consensus ($4.45 \pm 0.30$ versus $4.53 \pm 0.23$), with higher Evidence scores (4.57 versus 3.92). In prospective multicentre evaluation (59 patients), OMGs demonstrated high concordance with routine MDT decisions. Critically, in paired human-AI studies, OMGs most substantially enhanced clinicians' recommendations in Evidence and Robustness, the dimensions most compromised when multidisciplinary expertise is unavailable. These findings suggest that multi-agent deliberative systems can achieve performance comparable to expert MDT consensus, with potential to expand access to specialized oncology expertise in resource-limited settings.

Artificial intelligence built on large foundation models has transformed language understanding, vision and reasoning, yet these systems remain isolated and cannot readily share their capabilities. Integrating the complementary strengths of such independent foundation models is essential for building trustworthy intelligent systems. Despite rapid progress in individual model design, there is no established approach for coordinating such black-box heterogeneous models. Here we show that coordination can be achieved through a meta-ensemble framework termed StackingNet, which draws on principles of collective intelligence to combine model predictions during inference. StackingNet improves accuracy, reduces bias, enables reliability ranking, and identifies or prunes models that degrade performance, all operating without access to internal parameters or training data. Across tasks involving language comprehension, visual estimation, and academic paper rating, StackingNet consistently improves accuracy, robustness, and fairness, compared with individual models and classic ensembles. By turning diversity from a source of inconsistency into collaboration, StackingNet establishes a practical foundation for coordinated artificial intelligence, suggesting that progress may emerge from not only larger single models but also principled cooperation among many specialized ones.

Predicting transcriptional responses to unseen genetic perturbations is essential for understanding gene regulation and prioritizing large-scale perturbation experiments. Existing approaches either rely on static, potentially incomplete knowledge graphs, or prompt language models for functionally similar genes, retrieving associations shaped by symmetric co-occurrence in scientific text rather than directed regulatory logic. We introduce MechPert, a lightweight framework that encourages LLM agents to generate directed regulatory hypotheses rather than relying solely on functional similarity. Multiple agents independently propose candidate regulators with associated confidence scores; these are aggregated through a consensus mechanism that filters spurious associations, producing weighted neighborhoods for downstream prediction. We evaluate MechPert on Perturb-seq benchmarks across four human cell lines. For perturbation prediction in low-data regimes ($N=50$ observed perturbations), MechPert improves Pearson correlation by up to 10.5\% over similarity-based baselines. For experimental design, MechPert-selected anchor genes outperform standard network centrality heuristics by up to 46\% in well-characterized cell lines.

Sign language lexicographers construct bilingual dictionaries by establishing word-to-sign mappings, where polysemous and homonymous words corresponding to different signs across contexts are often underrepresented. A usage-based approach examining how word senses map to signs can identify such novel mappings absent from current dictionaries, enriching lexicographic resources. We address this by analyzing German and German Sign Language (Deutsche Gebärdensprache, DGS), manually annotating 1,404 word use-to-sign ID mappings derived from 32 words from the German Word Usage Graph (D-WUG) and 49 signs from the Digital Dictionary of German Sign Language (DW-DGS). We identify three correspondence types: Type 1 (one-to-many), Type 2 (many-to-one), and Type 3 (one-to-one), plus No Match cases. We evaluate computational methods: Exact Match (EM) and Semantic Similarity (SS) using SBERT embeddings. SS substantially outperforms EM overall 88.52% vs. 71.31%), with dramatic gains for Type 1 (+52.1 pp). Our work establishes the first annotated dataset for cross-modal sense correspondence and reveals which correspondence patterns are computationally identifiable. Our code and dataset are made publicly available.

As cloud computing scales toward the Exascale regime ($10^5+$ nodes), the prevailing "Newtonian" orchestration paradigm -- exemplified by Kubernetes -- approaches fundamental physical limits. The centralized, deterministic scheduling model suffers from $O(N)$ latency scaling, "Head-of-Line" blocking, and thermodynamic blindness, rendering it incapable of managing the stochastic chaos of next-generation AI workloads. This paper proposes a paradigm shift from orchestration to Thermodynamic Governance. We model the compute cluster not as a static state machine, but as a Dissipative Structure far from equilibrium. We introduce TEG (Thermo-Economic Governor), a decentralized architecture that establishes a rigorous topological isomorphism between cluster resource contention and many-body physics. TEG replaces the global scheduler with Langevin Agents that execute Brownian motion on a Holographic Potential Field, reducing decision complexity to $O(1)$. System stability is maintained via a macro-scale Landau Phase Transition mechanism, which modulates global damping (taxation) to physically dissolve deadlocks. Crucially, we enforce Token Evaporation to mirror entropy dissipation, preventing economic inflation and ensuring an open thermodynamic system. We provide formal theoretical analysis proving that: (1) The system converges asymptotically to a Nash Equilibrium via Dual-Number Damping; (2) OOM catastrophic failures are converted into manageable Glassy States via an OS-level Airlock Mutex; and (3) Safety is mathematically guaranteed under high inertia using High-Order Control Barrier Functions (HOCBF). TEG demonstrates that emergent order, rather than deterministic control, is the necessary condition for Exascale scalability.

Explaining with examples is an intuitive way to justify AI decisions. However, it is challenging to understand how a decision value should change relative to the examples with many features differing by large amounts. We draw from real estate valuation that uses Comparables-examples with known values for comparison. Estimates are made more accurate by hypothetically adjusting the attributes of each Comparable and correspondingly changing the value based on factors. We propose Comparables XAI for relatable example-based explanations of AI with Trace adjustments that trace counterfactual changes from each Comparable to the Subject, one attribute at a time, monotonically along the AI feature space. In modelling and user studies, Trace-adjusted Comparables achieved the highest XAI faithfulness and precision, user accuracy, and narrowest uncertainty bounds compared to linear regression, linearly adjusted Comparables, or unadjusted Comparables. This work contributes a new analytical basis for using example-based explanations to improve user understanding of AI decisions.

While Time Series Foundation Models (TSFMs) have demonstrated exceptional performance in generalized forecasting, their performance often degrades significantly when deployed in real-world vertical domains characterized by temporal distribution shifts and domain-specific periodic structures. Current solutions are primarily constrained by two paradigms: Domain-Adaptive Pretraining (DAPT), which improves short-term domain fitting but frequently disrupts previously learned global temporal patterns due to catastrophic forgetting; and Retrieval-Augmented Generation (RAG), which incorporates external knowledge but introduces substantial retrieval overhead. This creates a severe scalability bottleneck that fails to meet the high-efficiency requirements of real-time stream processing. To break this impasse, we propose Memory for Time Series (MEMTS), a lightweight and plug-and-play method for retrieval-free domain adaptation in time series forecasting. The key component of MEMTS is a Knowledge Persistence Module (KPM), which internalizes domain-specific temporal dynamics, such as recurring seasonal patterns and trends into a compact set of learnable latent prototypes. In doing so, it transforms fragmented historical observations into continuous, parameterized knowledge representations. This paradigm shift enables MEMTS to achieve accurate domain adaptation with constant-time inference and near-zero latency, while effectively mitigating catastrophic forgetting of general temporal patterns, all without requiring any architectural modifications to the frozen TSFM backbone. Extensive experiments on multiple datasets demonstrate the SOTA performance of MEMTS.

Context: The Impostor Phenomenon (IP), the persistent belief of being a fraud despite evident competence, is common in Software Engineering (SE), where high expectations for expertise and innovation prevail. Although coaching and similar interventions are proposed to mitigate IP, empirical evidence in SE remains underexplored. Objective: This study examines the impact of a structured group coaching intervention on reducing IP feelings among early-career software engineers. Method: We conducted a quasi-experiment with 20 participants distributed across two project teams using a wait-list control design, complemented by non-participant observation. The treatment group received a three-session coaching intervention, while the control group received it after an observation phase. IP was assessed using the Clance Impostor Phenomenon Scale (CIPS), alongside evaluated measures of well-being (WHO-5), life satisfaction (SWLS), and affect (PANAS). Results: The coaching resulted in modest reductions in CIPS scores, whereas the control group also improved during the observation phase, suggesting that contextual and temporal factors may have exerted a stronger influence than the formal intervention. Conclusion: These results suggest that coaching may support reflection and awareness related to IP, yet other contextual aspects of team collaboration and project work might also contribute to these changes. This study offers a novel empirical step toward understanding how structured IP interventions operate within SE environments.

Data quality is a crucial factor in large language models training. While prior work has shown that models trained on smaller, high-quality datasets can outperform those trained on much larger but noisy or low-quality corpora, systematic methods for industrial-scale data selection in instruction tuning remain underexplored. In this work, we study instruction-tuning data selection through the lens of semantic representation similarity and identify a key limitation of state-of-the-art LLM encoders: they produce highly redundant semantic embeddings. To mitigate this redundancy, we propose Compressed Representation Data Selection (CRDS), a novel framework with two variants. CRDS-R applies Rademacher random projection followed by concatenation of transformer hidden-layer representations, while CRDS-W employs whitening-based dimensionality reduction to improve representational quality. Experimental results demonstrate that both variants substantially enhance data quality and consistently outperform state-of-the-art representation-based selection methods. Notably, CRDS-W achieves strong performance using only 3.5% of the data, surpassing the full-data baseline by an average of 0.71% across four datasets. Our code is available at https://github.com/tdano1/CRDS.

Large Language Models (LLMs) have demonstrated strong semantic reasoning across multimodal domains. However, their integration with graph-based models of brain connectivity remains limited. In addition, most existing fMRI analysis methods rely on static Functional Connectivity (FC) representations, which obscure transient neural dynamics critical for neurodevelopmental disorders such as autism. Recent state-space approaches, including Mamba, model temporal structure efficiently, but are typically used as standalone feature extractors without explicit high-level reasoning. We propose NeuroMambaLLM, an end-to-end framework that integrates dynamic latent graph learning and selective state-space temporal modelling with LLMs. The proposed method learns the functional connectivity dynamically from raw Blood-Oxygen-Level-Dependent (BOLD) time series, replacing fixed correlation graphs with adaptive latent connectivity while suppressing motion-related artifacts and capturing long-range temporal dependencies. The resulting dynamic brain representations are projected into the embedding space of an LLM model, where the base language model remains frozen and lightweight low-rank adaptation (LoRA) modules are trained for parameter-efficient alignment. This design enables the LLM to perform both diagnostic classification and language-based reasoning, allowing it to analyze dynamic fMRI patterns and generate clinically meaningful textual reports.

Automating scientific discovery in complex, experiment-driven domains requires more than iterative mutation of programs; it demands structured hypothesis management, environment interaction, and principled reflection. We present OR-Agent, a configurable multi-agent research framework designed for automated exploration in rich experimental environments. OR-Agent organizes research as a structured tree-based workflow that explicitly models branching hypothesis generation and systematic backtracking, enabling controlled management of research trajectories beyond simple mutation-crossover loops. At its core, we introduce an evolutionary-systematic ideation mechanism that unifies evolutionary selection of research starting points, comprehensive research plan generation, and coordinated exploration within a research tree. We further propose a hierarchical optimization-inspired reflection system: short-term experimental reflection operates as a form of verbal gradient providing immediate corrective signals; long-term reflection accumulates cross-experiment insights as verbal momentum; and memory compression serves as a regularization mechanism analogous to weight decay, preserving essential signals while mitigating drift. Together, these components form a principled architecture governing research dynamics. We conduct extensive experiments across classical combinatorial optimization benchmarks-including traveling salesman, capacitated vehicle routing, bin packing, orienteering, and multiple knapsack problems-as well as simulation-based cooperative driving scenarios. Results demonstrate that OR-Agent outperforms strong evolutionary baselines while providing a general, extensible, and inspectable framework for AI-assisted scientific discovery. OR-Agent source code and experiments data are publicly available at https://github.com/qiliuchn/OR-Agent.

The Impostor Phenomenon (IP) impacts a significant portion of the Software Engineering workforce, yet it is often viewed primarily through an internal individual lens. In this position paper, we propose framing the prevalence of IP as a form of Human Debt and discuss the relation with the ICSE2026 Pre Survey on the Future of Software Engineering results. Similar to technical debt, which arises when short-term goals are prioritized over long-term structural integrity, Human Debt accumulates due to gaps in psychological safety and inclusive support within socio-technical ecosystems. We observe that this debt is not distributed equally, it weighs heavier on underrepresented engineers and researchers, who face compounded challenges within traditional hierarchical structures and academic environments. We propose cultural refactoring, transparency and active maintenance through allyship, suggesting that leaders and institutions must address the environmental factors that exacerbate these feelings, ensuring a sustainable ecosystem for all professionals.

Context: Generative Artificial Intelligence (GenAI) tools, such as GitHub Copilot and GPT tools, represent a paradigm shift in software engineering. While their impact is clear, most studies are short-term, focused on individual experiments. The sustained, team-level effects on productivity within industrial agile environments remain largely uncharacterized. Goal: This study aims to provide a longitudinal evaluation of GenAI's impact on agile software teams. We characterize its effect on developers' productivity by applying the multi-dimensional SPACE framework. Method: We conducted a multi-case longitudinal study involving 3 agile teams at a large technology consulting firm for around 13 months. We collected and compared quantitative telemetry (Jira, SonarQube, Git) and qualitative survey data from historical (pre-adoption) and research (post-adoption) sprints. Conclusion: GenAI tools can significantly improve team performance and well-being. Our key finding is a sharp increase in Performance and perceived Efficiency concurrent with flat developer Activity. This suggests GenAI increases the value density of development work, not its volume. This finding validates the necessity of multi-dimensional frameworks like SPACE to capture the true, nuanced impact of GenAI in situ, which would be invisible to studies measuring Activity alone.

While vision-language-action (VLA) models have advanced generalist robotic learning, cross-embodiment transfer remains challenging due to kinematic heterogeneity and the high cost of collecting sufficient real-world demonstrations to support fine-tuning. Existing cross-embodiment policies typically rely on shared-private architectures, which suffer from limited capacity of private parameters and lack explicit adaptation mechanisms. To address these limitations, we introduce MOTIF for efficient few-shot cross-embodiment transfer that decouples embodiment-agnostic spatiotemporal patterns, termed action motifs, from heterogeneous action data. Specifically, MOTIF first learns unified motifs via vector quantization with progress-aware alignment and embodiment adversarial constraints to ensure temporal and cross-embodiment consistency. We then design a lightweight predictor that predicts these motifs from real-time inputs to guide a flow-matching policy, fusing them with robot-specific states to enable action generation on new embodiments. Evaluations across both simulation and real-world environments validate the superiority of MOTIF, which significantly outperforms strong baselines in few-shot transfer scenarios by 6.5% in simulation and 43.7% in real-world settings. Code is available at https://github.com/buduz/MOTIF.

We study matrix-free eigendecomposition under a matrix-vector product (MVP) oracle, where each step observes a covariance operator $C_k = C_{sig} + σ_k^2 I + E_k$. Standard stochastic approximation methods either use fixed steps that couple stability to $\|C_k\|_2$, or adapt steps in ways that slow down due to vanishing updates. We introduce a discrete double-bracket flow whose generator is invariant to isotropic shifts, yielding pathwise invariance to $σ_k^2 I$ at the discrete-time level. The resulting trajectory and a maximal stable step size $η_{max} \propto 1/\|C_e\|_2^2$ depend only on the trace-free covariance $C_e$. We establish global convergence via strict-saddle geometry for the diagonalization objective and an input-to-state stability analysis, with sample complexity scaling as $O(\|C_e\|_2^2 / (Δ^2 ε))$ under trace-free perturbations. An explicit characterization of degenerate blocks yields an accelerated $O(\log(1/ζ))$ saddle-escape rate and a high-probability finite-time convergence guarantee.

Multimodal Large Language Models demonstrate strong performance on natural image understanding, yet exhibit limited capability in interpreting scientific images, including but not limited to schematic diagrams, experimental characterizations, and analytical charts. This limitation is particularly pronounced in open-source MLLMs. The gap largely stems from existing datasets with limited domain coverage, coarse structural annotations, and weak semantic grounding. We introduce OmniScience, a large-scale, high-fidelity multi-modal dataset comprising 1.5 million figure-caption-context triplets, spanning more than 10 major scientific disciplines. To obtain image caption data with higher information density and accuracy for multi-modal large-model training, we develop a dynamic model-routing re-captioning pipeline that leverages state-of-the-art multi-modal large language models to generate dense, self-contained descriptions by jointly synthesizing visual features, original figure captions, and corresponding in-text references authored by human scientists. The pipeline is further reinforced with rigorous quality filtering and alignment with human expert judgments, ensuring both factual accuracy and semantic completeness, and boosts the image-text multi-modal similarity score from 0.769 to 0.956. We further propose a caption QA protocol as a proxy task for evaluating visual understanding. Under this setting, Qwen2.5-VL-3B model finetuned on OmniScience show substantial gains over baselines, achieving a gain of 0.378 on MM-MT-Bench and a gain of 0.140 on MMMU.

Multimedia Event Extraction (MEE) aims to identify events and their arguments from documents that contain both text and images. It requires grounding event semantics across different modalities. Progress in MEE is limited by the lack of annotated training data. M2E2 is the only established benchmark, but it provides annotations only for evaluation. This makes direct supervised training impractical. Existing methods mainly rely on cross-modal alignment or inference-time prompting with Vision--Language Models (VLMs). These approaches do not explicitly learn structured event representations and often produce weak argument grounding in multimodal settings. To address these limitations, we propose RMPL, a Relation-aware Multi-task Progressive Learning framework for MEE under low-resource conditions. RMPL incorporates heterogeneous supervision from unimodal event extraction and multimedia relation extraction with stage-wise training. The model is first trained with a unified schema to learn shared event-centric representations across modalities. It is then fine-tuned for event mention identification and argument role extraction using mixed textual and visual data. Experiments on the M2E2 benchmark with multiple VLMs show consistent improvements across different modality settings.

Industrial thermal power systems have coupled performance variables with hierarchical order of importance, making their simultaneous optimization computationally challenging or infeasible. This barrier limits the integrated and computationally scaleable operation optimization of industrial thermal power systems. To address this issue for large-scale engineering systems, we present a fully machine learning-powered bi-level optimization framework for data-driven optimization of industrial thermal power systems. The objective functions of upper and lower levels are approximated by artificial neural network (ANN) models and the lower-level problem is analytically embedded through Karush-Kuhn-Tucker (KKT) optimality conditions. The reformulated single level optimization framework integrating ANN models and KKT constraints (ANN-KKT) is validated on benchmark problems and on real-world power generation operation of 660 MW coal power plant and 395 MW gas turbine system. The results reveal a comparable solutions obtained from the proposed ANN-KKT framework to the bi-level solutions of the benchmark problems. Marginal computational time requirement (0.22 to 0.88 s) to compute optimal solutions yields 583 MW (coal) and 402 MW (gas turbine) of power output at optimal turbine heat rate of 7337 kJ/kWh and 7542 kJ/kWh, respectively. In addition, the method expands to delineate a feasible and robust operating envelope that accounts for uncertainty in operating variables while maximizing thermal efficiency in various scenarios. These results demonstrate that ANN-KKT offers a scalable and computationally efficient route for hierarchical, data-driven optimization of industrial thermal power systems, achieving energy-efficient operations of large-scale engineering systems and contributing to industry 5.0.

Chain-of-thought (CoT) prompting improves reasoning but often increases inference cost by one to two orders of magnitude. To address these challenges, we present \textbf{OneLatent}, a framework that compresses intermediate reasoning into a single latent token via supervision from rendered CoT images and DeepSeek-OCR hidden states. By rendering textual steps into images, we obtain a deterministic supervision signal that can be inspected and audited without requiring the model to output verbose textual rationales. Across benchmarks, OneLatent reduces average output length by $11\times$ with only a $2.21\%$ average accuracy drop relative to textual CoT, while improving output token contribution (OTC) by $6.8\times$. On long-chain logical reasoning, OneLatent reaches $99.80\%$ on ProntoQA and $97.80\%$ on ProsQA with one latent token, with compression up to $87.4\times$, supporting compression-constrained generalization.

Quality-Diversity (QD) algorithms aim to discover diverse, high-performing solutions across behavioral niches. However, QD search often stagnates as incremental variation operators struggle to propagate building blocks across large populations. Existing mutation operators rely on gradual variation to solutions, limiting their ability to efficiently explore regions of the search space distant from parent solutions or to spread beneficial genetic material through the population. We propose a mutation operator which augments variation-based operators with discrete, gene-level crossover, enabling rapid recombination of elite genetic material. This crossover mechanism mirrors the biological principle of meiosis and facilitates both the direct transfer of genetic material and the exploration of novel genotype configurations beyond the existing elite hypervolume. We evaluate operators on three locomotion environments, demonstrating improvements in QD score, coverage, and max fitness, with particularly strong performance in later stages of optimization once building blocks have been established in the archive. These results show that the addition of a discrete crossover mutation provides a complementary exploration mechanism that sustains quality-diversity growth beyond the performance demonstrated by existing operators.

Large Language Models (LLMs) have improved programming efficiency, but their performance degrades significantly as requirements scale; when faced with multi-modal documents containing hundreds of scenarios, LLMs often produce incorrect implementations or omit constraints. We propose Agentic Requirement Compilation (ARC), a technique that moves beyond simple code generation to requirement compilation, enabling the creation of runnable web systems directly from multi-modal DSL documents. ARC generates not only source code but also modular designs for UI, API, and database layers, enriched test suites (unit, modular, and integration), and detailed traceability for software maintenance. Our approach employs a bidirectional test-driven agentic loop: a top-down architecture phase decomposes requirements into verifiable interfaces, followed by a bottom-up implementation phase where agents generate code to satisfy those tests. ARC maintains strict traceability across requirements, design, and code to facilitate intelligent asset reuse. We evaluated ARC by generating six runnable web systems from documents spanning 50-200 multi-modal scenarios. Compared to state-of-the-art baselines, ARC-generated systems pass 50.6% more GUI tests on average. A user study with 21 participants showed that novice users can successfully write DSL documents for complex systems, such as a 10K-line ticket-booking system, in an average of 5.6 hours. These results demonstrate that ARC effectively transforms non-trivial requirement specifications into maintainable, runnable software.

Limited by inference latency, existing robot manipulation policies lack sufficient real-time interaction capability with the environment. Although faster generation methods such as flow matching are gradually replacing diffusion methods, researchers are pursuing even faster generation suitable for interactive robot control. MeanFlow, as a one-step variant of flow matching, has shown strong potential in image generation, but its precision in action generation does not meet the stringent requirements of robotic manipulation. We therefore propose \textbf{HybridFlow}, a \textbf{3-stage method} with \textbf{2-NFE}: Global Jump in MeanFlow mode, ReNoise for distribution alignment, and Local Refine in ReFlow mode. This method balances inference speed and generation quality by leveraging the rapid advantage of MeanFlow one-step generation while ensuring action precision with minimal generation steps. Through real-world experiments, HybridFlow outperforms the 16-step Diffusion Policy by \textbf{15--25\%} in success rate while reducing inference time from 152ms to 19ms (\textbf{8$\times$ speedup}, \textbf{$\sim$52Hz}); it also achieves 70.0\% success on unseen-color OOD grasping and 66.3\% on deformable object folding. We envision HybridFlow as a practical low-latency method to enhance real-world interaction capabilities of robotic manipulation policies.

Identifying the strategic uses of reformulation in discourse remains a key challenge for computational argumentation. While LLMs can detect surface-level similarity, they often fail to capture the pragmatic functions of rephrasing, such as its role within rhetorical discourse. This paper presents a comparative multi-agent framework designed to quantify the benefits of incorporating explicit theoretical knowledge for this task. We utilise an dataset of annotated political debates to establish a new standard encompassing four distinct rephrase functions: Deintensification, Intensification, Specification, Generalisation, and Other, which covers all remaining types (D-I-S-G-O). We then evaluate two parallel LLM-based agent systems: one enhanced by argumentation theory via Retrieval-Augmented Generation (RAG), and an identical zero-shot baseline. The results reveal a clear performance gap: the RAG-enhanced agents substantially outperform the baseline across the board, with particularly strong advantages in detecting Intensification and Generalisation context, yielding an overall Macro F1-score improvement of nearly 30\%. Our findings provide evidence that theoretical grounding is not only beneficial but essential for advancing beyond mere paraphrase detection towards function-aware analysis of argumentative discourse. This comparative multi-agent architecture represents a step towards scalable, theoretically informed computational tools capable of identifying rhetorical strategies in contemporary discourse.

Soil-transmitted helminth (STH) infections continuously affect a large proportion of the global population, particularly in tropical and sub-tropical regions, where access to specialized diagnostic expertise is limited. Although manual microscopic diagnosis of parasitic eggs remains the diagnostic gold standard, the approach can be labour-intensive, time-consuming, and prone to human error. This paper aims to utilize a vision language model (VLM) such as Microsoft Florence that was fine-tuned to localize all parasitic eggs within microscopic images. The preliminary results show that our localization VLM performs comparatively better than the other object detection methods, such as EfficientDet, with an mIOU of 0.94. This finding demonstrates the potential of the proposed VLM to serve as a core component of an automated framework, offering a scalable engineering solution for intelligent parasitological diagnosis.

Vision-Language-Action (VLA) models enable instruction-following embodied control, but their large compute and memory footprints hinder deployment on resource-constrained robots and edge platforms. While reducing weights to 1-bit precision through binarization can greatly improve efficiency, existing methods fail to narrow the distribution gap between binarized and full-precision weights, causing quantization errors to accumulate under long-horizon closed-loop execution and severely degrade actions. To fill this gap, we propose HBVLA, a VLA-tailored binarization framework. First, we use a policy-aware enhanced Hessian to identify weights that are truly critical for action generation. Then, we employ a sparse orthogonal transform for non-salient weights to induce a low-entropy intermediate state. Finally, we quantize both salient and non-salient weights in the Harr domain with group-wise 1-bit quantization. We have evaluated our approach on different VLAs: on LIBERO, quantized OpenVLA-OFT retains 92.2% of full-precision performance; on SimplerEnv, quantized CogAct retains 93.6%, significantly outperforming state-of-the-art binarization methods. We further validate our method on real-world evaluation suite and the results show that HBVLA incurs only marginal success-rate degradation compared to the full-precision model, demonstrating robust deployability under tight hardware constraints. Our work provides a practical foundation for ultra-low-bit quantization of VLAs, enabling more reliable deployment on hardware-limited robotic platforms.

We introduce \texttt{OPO-CMDP}, the first policy optimization algorithm for stochastic Contextual Markov Decision Process (CMDPs) under general offline function approximation. Our approach achieves a high probability regret bound of $\widetilde{O}(H^4\sqrt{T|S||A|\log(|\mathcal{F}||\mathcal{P}|)}),$ where $S$ and $A$ denote the state and action spaces, $H$ the horizon length, $T$ the number of episodes, and $\mathcal{F}, \mathcal{P}$ the finite function classes used to approximate the losses and dynamics, respectively. This is the first regret bound with optimal dependence on $|S|$ and $|A|$, directly improving the current state-of-the-art (Qian, Hu, and Simchi-Levi, 2024). These results demonstrate that optimistic policy optimization provides a natural, computationally superior and theoretically near-optimal path for solving CMDPs.

In this work, we presented Pailitao-VL, a comprehensive multi-modal retrieval system engineered for high-precision, real-time industrial search. We here address three critical challenges in the current SOTA solution: insufficient retrieval granularity, vulnerability to environmental noise, and prohibitive efficiency-performance gap. Our primary contribution lies in two fundamental paradigm shifts. First, we transitioned the embedding paradigm from traditional contrastive learning to an absolute ID-recognition task. Through anchoring instances to a globally consistent latent space defined by billions of semantic prototypes, we successfully overcome the stochasticity and granularity bottlenecks inherent in existing embedding solutions. Second, we evolved the generative reranker from isolated pointwise evaluation to the compare-and-calibrate listwise policy. By synergizing chunk-based comparative reasoning with calibrated absolute relevance scoring, the system achieves nuanced discriminative resolution while circumventing the prohibitive latency typically associated with conventional reranking methods. Extensive offline benchmarks and online A/B tests on Alibaba e-commerce platform confirm that Pailitao-VL achieves state-of-the-art performance and delivers substantial business impact. This work demonstrates a robust and scalable path for deploying advanced MLLM-based retrieval architectures in demanding, large-scale production environments.

Metaphors are a distinctive feature of literary language, yet they remain less studied experimentally than everyday metaphors. Moreover, previous psycholinguistic and computational approaches overlooked the temporal dimension, although many literary metaphors were coined centuries apart from contemporary readers. This study innovatively applies tools from diachronic distributional semantics to assess whether the processing costs of literary metaphors varied over time and genre. Specifically, we trained word embeddings on literary and nonliterary Italian corpora from the 19th and 21st centuries, for a total of 124 million tokens, and modeled changes in the semantic similarity between topics and vehicles of 515 19th-century literary metaphors, taking this measure as a proxy of metaphor processing demands. Overall, semantic similarity, and hence metaphor processing demands, remained stable over time. However, genre played a key role: metaphors appeared more difficult (i.e., lower topic-vehicle similarity) in modern literary contexts than in 19th-century literature, but easier (i.e., higher topic-vehicle similarity) in today's nonliterary language (e.g., the Web) than in 19th-century nonliterary texts. This pattern was further shaped by semantic features of metaphors' individual terms, such as vector coherence and semantic neighborhood density. Collectively, these findings align with broader linguistic changes in Italian, such as the stylistic simplification of modern literature, which may have increased metaphor processing demands, and the high creativity of the Web's language, which seems to render metaphor more accessible.

We present the first high-probability optimal regret bound for a policy optimization technique applied to the problem of stochastic contextual multi-armed bandit (CMAB) with general offline function approximation. Our algorithm is both efficient and achieves an optimal regret bound of $\widetilde{O}(\sqrt{ K|\mathcal{A}|\log|\mathcal{F}|})$, where $K$ is the number of rounds, $\mathcal{A}$ is the set of arms, and $\mathcal{F}$ is the function class used to approximate the losses. Our results bridge the gap between theory and practice, demonstrating that the widely used policy optimization methods for the contextual bandit problem can achieve a rigorously-proved optimal regret bound. We support our theoretical results with an empirical evaluation of our algorithm.

Large language models (LLMs) often generate plausible yet incorrect answers, posing risks in safety-critical settings such as medicine. Human evaluation is expensive, and LLM-as-judge approaches risk introducing hidden errors. Recent white-box methods detect contextual hallucinations using model internals, focusing on the localization of the attention mass, but two questions remain open: do these approaches extend to predicting answer correctness, and do they generalize out-of-domains? We introduce Head Entropy, a method that predicts answer correctness from attention entropy patterns, specifically measuring the spread of the attention mass. Using sparse logistic regression on per-head 2-Renyi entropies, Head Entropy matches or exceeds baselines in-distribution and generalizes substantially better on out-of-domains, it outperforms the closest baseline on average by +8.5% AUROC. We further show that attention patterns over the question/context alone, before answer generation, already carry predictive signal using Head Entropy with on average +17.7% AUROC over the closest baseline. We evaluate across 5 instruction-tuned LLMs and 3 QA datasets spanning general knowledge, multi-hop reasoning, and medicine.

Relational databases (RDBs) contain vast amounts of heterogeneous tabular information that can be exploited for predictive modeling purposes. But since the space of potential targets is vast across enterprise settings, how can we \textit{avoid retraining} a new model each time we wish to predict a new quantity of interest? Foundation models based on in-context learning (ICL) offer a convenient option, but so far are largely restricted to single-table operability. In generalizing to multiple interrelated tables, it is essential to compress variably-sized RDB neighborhoods into fixed-length ICL samples for consumption by the decoder. However, the details here are critical: unlike existing supervised learning RDB pipelines, we provide theoretical and empirical evidence that ICL-specific compression should be constrained \emph{within} high-dimensional RDB columns where all entities share units and roles, not \textit{across} columns where the relevance of heterogeneous data types cannot possibly be determined without label information. Conditioned on this restriction, we then demonstrate that encoder expressiveness is actually not compromised by excluding trainable parameters. Hence we arrive at a principled family of RDB encoders that can be seamlessly paired with already-existing single-table ICL foundation models, whereby no training or fine-tuning is required. From a practical standpoint, we develop scalable SQL primitives to implement the encoder stage, resulting in an easy-to-use open-source RDB foundation model\footnote{\label{foot: RDBLearn_learn} https://github.com/HKUSHXLab/rdblearn} capable of robust performance on unseen datasets out of the box.

Large language models (LLMs) have recently achieved remarkable success in generating rigorous mathematical proofs, with "AI for Math" emerging as a vibrant field of research. While these models have mastered competition-level benchmarks like the International Mathematical Olympiad and show promise in research applications through auto-formalization, their deployment via lightweight, natural-language pipelines for research problems remains underexplored. In this work, we demonstrate that next-generation models (e.g., Gemini 3 Pro, GPT-5.2 Pro), when integrated into a streamlined automated pipeline optimized for citation-based verification, can solve sophisticated research-grade problems. We evaluate our pipeline on two novel datasets: (1) the ICCM problem sets (comparable to the S.-T. Yau College Student Mathematics Contest) proposed by leading mathematicians, and (2) the "First Proof" problem set, consisting of previously unpublished research questions. Our pipeline generated candidate proofs for all problems in the first two ICCM sets and the "First Proof" set. The solutions for the first two ICCM sets and Problem 4 of the "First Proof" set have been fully verified by our team. All generated proofs have been submitted to the official organization, and our generated results are publicly available. We plan to open-source the complete pipeline methodology in due course.

Large language models(LLMs) are now used to power complex multi-turn agentic workflows. Existing systems run agentic inference by loosely assembling isolated components: an LLM inference engine (e.g., vLLM) and a tool orchestrator (e.g., Kubernetes). Although agentic workflows involve multiple LLM and tool requests, these systems schedule and allocate resources separately on a per-request basis, without end-to-end knowledge of the workflow. This leads to sub-optimal management of KV cache and tool execution environments. To address the challenges, we propose ThunderAgent, a fast, simple, and program-aware agentic inference system. We first abstract agentic workflows as LLM Programs, enabling a unified view of heterogeneous resources, including KV caches, system states, and external tool assets such as disk memory and network ports. Built upon this abstraction, ThunderAgent introduces a program-aware scheduler and a tool resource manager designed to maximize KV cache hit rates, mitigate memory imbalances, and enable asynchronous environment preparation. Evaluations across coding, routing, and scientific discovery agents demonstrate that ThunderAgent achieves 1.5-3.6x throughput improvements in serving, 1.8-3.9x in RL rollout, and up to 4.2x disk memory savings compared to state-of-the-art inference systems. To facilitate reproducibility and support future development, we open-source the system implementations of the whole ThunderAgent at: https://github.com/Agentic-Kinetics/ThunderAgent.

Recent advancements in Large Language Model (LLM) agents have demonstrated strong capabilities in executing complex tasks through tool use. However, long-horizon multi-step tool planning is challenging, because the exploration space suffers from a combinatorial explosion. In this scenario, even when a correct tool-use path is found, it is usually considered an immediate reward for current training, which would not provide any reusable information for subsequent training. In this paper, we argue that historically successful trajectories contain reusable tool-transition patterns, which can be leveraged throughout the whole training process. Inspired by ant colony optimization where historically successful paths can be reflected by the pheromone, we propose Pheromone-Guided Policy Optimization (PhGPO), which learns a trajectory-based transition pattern (i.e., pheromone) from historical trajectories and then uses the learned pheromone to guide policy optimization. This learned pheromone provides explicit and reusable guidance that steers policy optimization toward historically successful tool transitions, thereby improving long-horizon tool planning. Comprehensive experimental results demonstrate the effectiveness of our proposed PhGPO.

Magnetic-anomaly navigation, leveraging small-scale variations in the Earth's magnetic field, is a promising alternative when GPS is unavailable or compromised. Airborne systems face a key challenge in extracting geomagnetic field data: the aircraft itself induces magnetic noise. Although the classical Tolles-Lawson model addresses this, it inadequately handles stochastically corrupted magnetic data required for navigation. To address stochastic noise, we propose a framework based on two physics-based constraints: divergence-free vector field and E(3)-equivariance. These ensure the learned magnetic field obeys Maxwell's equations and that outputs transform correctly with sensor position/orientation. The divergence-free constraint is implemented by training a neural network to output a vector potential $A$, with the magnetic field defined as its curl. For E(3)-equivariance, we use tensor products of geometric tensors representable via spherical harmonics with known rotational transformations. Enforcing physical consistency and restricting the admissible function space acts as an implicit regularizer that improves spatio-temporal performance. We present ablation studies evaluating each constraint alone and jointly across CNNs, MLPs, Liquid Time Constant models, and Contiformers. Continuous-time dynamics and long-term memory are critical for modelling magnetic time series; the Contiformer architecture, which provides both, outperforms state-of-the-art methods. To mitigate data scarcity, we generate synthetic datasets using the World Magnetic Model (WMM) with time-series conditional GANs, producing realistic, temporally consistent magnetic sequences across varied trajectories and environments. Experiments show that embedding these constraints significantly improves predictive accuracy and physical plausibility, outperforming classical and unconstrained deep learning approaches.

Large Audio Language Models (LALMs) excel at perception but struggle with complex reasoning requiring precise acoustic measurements. While external tools can extract fine-grained features like exact tempo or pitch, effective integration remains challenging: naively using all tools causes information overload, while prompt-based selection fails to assess context-dependent utility. To address this, we propose AuTAgent (Audio Tool Agent), a reinforcement learning framework that learns when and which tools to invoke. By employing a sparse-feedback training strategy with a novel Differential Reward mechanism, the agent learns to filter out irrelevant tools and invokes external assistance only when it yields a net performance gain over the base model. Experimental results confirm that AuTAgent complements the representation bottleneck of LALMs by providing verifiable acoustic evidence. It improves accuracy by 4.20% / 6.20% and 9.80% / 8.00% for open-source and closed-source backbones on the MMAU Test-mini and the MMAR benchmarks, respectively. In addition, further experiments demonstrate exceptional transferability. We highlight the complementary role of external tools in augmenting audio model reasoning.

Correlation Clustering (CC) is a fundamental unsupervised learning primitive whose strongest LP-based approximation guarantees require $Θ(n^3)$ triangle inequality constraints and are prohibitive at scale. We initiate the study of \emph{sparsification--approximation trade-offs} for CC, asking how much edge information is needed to retain LP-based guarantees. We establish a structural dichotomy between pseudometric and general weighted instances. On the positive side, we prove that the VC dimension of the clustering disagreement class is exactly $n{-}1$, yielding additive $\varepsilon$-coresets of optimal size $\tilde{O}(n/\varepsilon^2)$; that at most $\binom{n}{2}$ triangle inequalities are active at any LP vertex, enabling an exact cutting-plane solver; and that a sparsified variant of LP-PIVOT, which imputes missing LP marginals via triangle inequalities, achieves a robust $\frac{10}{3}$-approximation (up to an additive term controlled by an empirically computable imputation-quality statistic $\overlineΓ_w$) once $\tildeΘ(n^{3/2})$ edges are observed, a threshold we prove is sharp. On the negative side, we show via Yao's minimax principle that without pseudometric structure, any algorithm observing $o(n)$ uniformly random edges incurs an unbounded approximation ratio, demonstrating that the pseudometric condition governs not only tractability but also the robustness of CC to incomplete information.

A Software Bill of Materials (SBOM) is a key component for the transparency of software supply chain; it is a structured inventory of the components, dependencies, and associated metadata of a software artifact. However, an SBOM often contain sensitive information that organizations are unwilling to disclose in full to anyone, for two main concerns: technological risks deriving from exposing proprietary dependencies or unpatched vulnerabilities, and business risks, deriving from exposing architectural strategies. Therefore, delivering a plaintext SBOM may result in the disruption of the intellectual property of a company. To address this, we present VeriSBOM, a trustless, selectively disclosed SBOM framework that provides cryptographic verifiability of SBOMs using zero-knowledge proofs. Within VeriSBOM, third parties can validate specific statements about a delivered software. Respectively, VeriSBOM allows independent third parties to verify if a software contains authentic dependencies distributed by official package managers and that the same dependencies satisfy rigorous policy constraints such as the absence of vulnerable dependencies or the adherence with specific licenses models. VeriSBOM leverages a scalable vector commitment scheme together with folding-based proof aggregation to produce succinct zero-knowledge proofs that attest to security and compliance properties while preserving confidentiality. Crucially, the verification process requires no trust in the SBOM publisher beyond the soundness of the underlying primitives, and third parties can independently check proofs against the public cryptographic commitments. We implement VeriSBOM, analyze its security, and evaluate its performance on real-world package registries. The results show that our method enables scalable, privacy-preserving, and verifiable SBOM sharing and validation.

Environmental pollution is a critical global issue, with recycling emerging as one of the most viable solutions. This study focuses on waste segregation, a crucial step in recycling processes to obtain raw material. Recent advancements in computer vision have significantly contributed to waste classification and recognition. In waste segregation, segmentation masks are essential for robots to accurately localize and pick objects from conveyor belts. The complexity of real-world waste environments, characterized by deformed items without specific patterns and overlapping objects, further complicates waste segmentation tasks. This paper proposes an Ensemble Learning approach to improve segmentation accuracy by combining high performing segmentation models, U-Net and FPN, using a weighted average method. U-Net excels in capturing fine details and boundaries in segmentation tasks, while FPN effectively handles scale variation and context in complex environments, and their combined masks result in more precise predictions. The dataset used closely mimics real-life waste scenarios, and preprocessing techniques were applied to enhance feature learning for deep learning segmentation models. The ensemble model, referred to as EL-4, achieved an IoU value of 0.8306, an improvement over U-Net's 0.8065, and reduced Dice loss to 0.09019 from FPN's 0.1183. This study could contribute to the efficiency of waste sorting at Material Recovery Facility, facilitating better raw material acquisition for recycling with minimal human intervention and enhancing the overall throughput.

Large Language Models (LLMs) encounter significant performance bottlenecks in long-sequence tasks due to the computational complexity and memory overhead inherent in the self-attention mechanism. To address these challenges, we introduce \textsc{AllMem}, a novel and efficient hybrid architecture that integrates Sliding Window Attention (SWA) with non-linear Test-Time Training (TTT) memory networks. \textsc{AllMem} enables models to effectively scale to ultra-long contexts while mitigating catastrophic forgetting. This approach not only overcomes the representation constraints typical of linear memory models but also significantly reduces the computational and memory footprint during long-sequence inference. Furthermore, we implement a Memory-Efficient Fine-Tuning strategy to replace standard attention layers in pre-trained models with memory-augmented sliding window layers. This framework facilitates the efficient transformation of any off-the-shelf pre-trained LLM into an \textsc{AllMem}-based architecture. Empirical evaluations confirm that our 4k window model achieves near-lossless performance on 37k LongBench with a marginal 0.83 drop compared to full attention. Furthermore, on InfiniteBench at a 128k context, our 8k window variant outperforms full attention, which validates the effectiveness of our parameterized memory in mitigating noise and maintaining robust long-range modeling without the prohibitive costs of global attention.

Current Explainable AI (XAI) focuses on explaining a single application, but when encountering related applications, users may rely on their prior understanding from previous explanations. This leads to either overgeneralization and AI overreliance, or burdensome independent memorization. Indeed, related decision tasks can share explanatory factors, but with some notable differences; e.g., body mass index (BMI) affects the risks for heart disease and diabetes at the same rate, but chest pain is more indicative of heart disease. Similarly, models using different attributes for the same task still share signals; e.g., temperature and pressure affect air pollution but in opposite directions due to the ideal gas law. Leveraging transfer of learning, we propose Transferable XAI to enable users to transfer understanding across related domains by explaining the relationship between domain explanations using a general affine transformation framework applied to linear factor explanations. The framework supports explanation transfer across various domain types: translation for data subspace (subsuming prior work on Incremental XAI), scaling for decision task, and mapping for attributes. Focusing on task and attributes domain types, in formative and summative user studies, we investigated how well participants could understand AI decisions from one domain to another. Compared to single-domain and domain-independent explanations, Transferable XAI was the most helpful for understanding the second domain, leading to the best decision faithfulness, factor recall, and ability to relate explanations between domains. This framework contributes to improving the reusability of explanations across related AI applications by explaining factor relationships between subspaces, tasks, and attributes.

Intent-Based Networking (IBN) simplifies network management, but its reliability is challenged by "intent drift", where the network's state gradually deviates from its intended goal, often leading to silent failures. Conventional approaches struggle to detect the subtle, early stages of intent drift, raising alarms only when degradation is significant and failure is imminent, which limits their effectiveness for proactive assurance. To address this, we propose LEAD-Drift, a framework that detects intent drift in real time to enable proactive failure prevention. LEAD-Drift's core contribution is reformulating intent failure detection as a supervised learning problem by training a lightweight neural network on fixed-horizon labels to predict a future risk score. The model's raw output is then smoothed with an Exponential Moving Average (EMA) and passed through a statistically tuned threshold to generate robust, real-time alerts. Furthermore, we enhance the framework with two key features for operational intelligence: a multi-horizon modeling technique for dynamic time-to-failure estimation, and per-alert explainability using SHAP to identify root-cause KPIs. Our evaluation on a time-series dataset shows LEAD-Drift provides significantly earlier warnings, improving the average lead time by 7.3 minutes (+17.8\%) compared to a distance-based baseline. It also reduces alert noise by 80.2\% compared to a weighted-KPI heuristic, with only a minor trade-off in lead time. These results demonstrate that LEAD-Drift as a highly effective, interpretable, and operationally efficient solution for proactive network assurance in IBN.

Large Language Model (LLM)-based multi-agent systems (MAS) have emerged as a promising paradigm for solving complex tasks. However, existing works often rely on manual designs or "one-size-fits-all" automation, lacking dynamic adaptability after deployment. Inspired by how biological systems adapt, we introduce MASFly, a novel multi-agent framework enabling dynamic adaptation at test time. To adapt system generation, MASFly employs a retrieval-augmented SOP instantiation mechanism that leverages a self-constructed repository of successful collaboration patterns, enabling the LLM to assemble customized MASs for new queries. For adaptive execution, MASFly incorporates an experience-guided supervision mechanism, where a dedicated Watcher agent monitors system behaviors with reference to a personalized experience pool and provides real-time interventions. Extensive experiments demonstrate that MASFly achieves state-of-the-art performance, most notably a 61.7% success rate on the TravelPlanner benchmark, while exhibiting strong task adaptability and robustness.

Class-incremental learning (CIL) with pre-trained models (PTMs) faces a critical trade-off between efficient adaptation and long-term stability. While analytic learning enables rapid, recursive closed-form updates, its efficacy is often compromised by accumulated errors and feature incompatibility. In this paper, we first conduct a systematic study to dissect the failure modes of PTM-based analytic CIL, identifying representation rigidity as the primary bottleneck. Motivated by these insights, we propose \textbf{VILA}, a novel dual-branch framework that advances analytic CIL via a two-level vision-language calibration strategy. Specifically, we coherently fuse plastic, task-adapted features with a frozen, universal semantic anchor at the feature level through geometric calibration, and leverage cross-modal priors at the decision level to rectify prediction bias. This confluence maintains analytic-learning's extreme efficiency while overcoming its inherent brittleness. Extensive experiments across eight benchmarks demonstrate that VILA consistently yields superior performance, particularly in fine-grained and long-sequence scenarios. Our framework harmonizes high-fidelity prediction with the simplicity of analytic learning. Our code is available at https://github.com/byzhaoAI/VILA

In complex clinical decision-making, clinicians must often track a variety of competing metrics defined by aim (ideal) and limit (strict) thresholds. Sifting through these high-dimensional tradeoffs to infer the optimal patient-specific strategy is cognitively demanding and historically prone to variability. In this paper, we address this challenge within the context of High-Dose-Rate (HDR) brachytherapy for cervical cancer, where planning requires strictly managing radiation hot spots while balancing tumor coverage against organ sparing. We present ALMo (Aim-Limit-defined Multi-Objective system), an interactive decision support system designed to infer and operationalize clinician intent. ALMo employs a novel optimization framework that minimizes manual input through automated parameter setup and enables flexible control over toxicity risks. Crucially, the system allows clinicians to navigate the Pareto surface of dosimetric tradeoffs by directly manipulating intuitive aim and limit values. In a retrospective evaluation of 25 clinical cases, ALMo generated treatment plans that consistently met or exceeded manual planning quality, with 65% of cases demonstrating dosimetric improvements. Furthermore, the system significantly enhanced efficiency, reducing average planning time to approximately 17 minutes, compared to the conventional 30-60 minutes. While validated in brachytherapy, ALMo demonstrates a generalized framework for streamlining interaction in multi-criteria clinical decision-making.

While agentic AI systems rely on LLMs to translate user intent into structured function calls, this process is fraught with computational redundancy, leading to high inference latency that hinders real-time applications. This paper identifies and addresses three key redundancies: (1) the redundant processing of a large library of function descriptions for every request; (2) the redundant use of a large, slow model to generate an entire, often predictable, token sequence; and (3) the redundant generation of fixed, boilerplate parameter syntax. We introduce HyFunc, a novel framework that systematically eliminates these inefficiencies. HyFunc employs a hybrid-model cascade where a large model distills user intent into a single "soft token." This token guides a lightweight retriever to select relevant functions and directs a smaller, prefix-tuned model to generate the final call, thus avoiding redundant context processing and full-sequence generation by the large model. To eliminate syntactic redundancy, our "dynamic templating" technique injects boilerplate parameter syntax on-the-fly within an extended vLLM engine. To avoid potential limitations in generalization, we evaluate HyFunc on an unseen benchmark dataset, BFCL. Experimental results demonstrate that HyFunc achieves an excellent balance between efficiency and performance. It achieves an inference latency of 0.828 seconds, outperforming all baseline models, and reaches a performance of 80.1%, surpassing all models with a comparable parameter scale. These results suggest that HyFunc offers a more efficient paradigm for agentic AI. Our code is publicly available at https://github.com/MrBlankness/HyFunc.

Foundation models and vision-language pre-training have significantly advanced Vision-Language Models (VLMs), enabling multimodal processing of visual and linguistic data. However, their application in domain-specific agricultural tasks, such as plant pathology, remains limited due to the lack of large-scale, comprehensive multimodal image--text datasets and benchmarks. To address this gap, we introduce LeafNet, a comprehensive multimodal dataset, and LeafBench, a visual question-answering benchmark developed to systematically evaluate the capabilities of VLMs in understanding plant diseases. The dataset comprises 186,000 leaf digital images spanning 97 disease classes, paired with metadata, generating 13,950 question-answer pairs spanning six critical agricultural tasks. The questions assess various aspects of plant pathology understanding, including visual symptom recognition, taxonomic relationships, and diagnostic reasoning. Benchmarking 12 state-of-the-art VLMs on our LeafBench dataset, we reveal substantial disparity in their disease understanding capabilities. Our study shows performance varies markedly across tasks: binary healthy--diseased classification exceeds 90\% accuracy, while fine-grained pathogen and species identification remains below 65\%. Direct comparison between vision-only models and VLMs demonstrates the critical advantage of multimodal architectures: fine-tuned VLMs outperform traditional vision models, confirming that integrating linguistic representations significantly enhances diagnostic precision. These findings highlight critical gaps in current VLMs for plant pathology applications and underscore the need for LeafBench as a rigorous framework for methodological advancement and progress evaluation toward reliable AI-assisted plant disease diagnosis. Code is available at https://github.com/EnalisUs/LeafBench.

In safety-critical applications such as medical image segmentation, prediction systems must provide reliability guarantees that extend beyond conventional expected loss control. While risk-controlling prediction sets (RCPS) offer probabilistic guarantees on the expected risk, they fail to capture tail behavior and worst-case scenarios that are crucial in high-stakes settings. This paper introduces optimized certainty equivalent RCPS (OCE-RCPS), a novel framework that provides high-probability guarantees on general optimized certainty equivalent (OCE) risk measures, including conditional value-at-risk (CVaR) and entropic risk. OCE-RCPS leverages upper confidence bounds to identify prediction set parameters that satisfy user-specified risk tolerance levels with provable reliability. We establish theoretical guarantees showing that OCE-RCPS satisfies the desired probabilistic constraint for loss functions such as miscoverage and false negative rate. Experiments on image segmentation demonstrate that OCE-RCPS consistently meets target satisfaction rates across various risk measures and reliability configurations, while OCE-CRC fails to provide probabilistic guarantees.

Fine-tuning large pretrained language models (LLMs) is a cornerstone of modern NLP, yet its growing memory demands (driven by backpropagation and large optimizer States) limit deployment in resource-constrained settings. Zero-order (ZO) methods bypass backpropagation by estimating directional derivatives from forward evaluations, offering substantial memory savings. However, classical ZO estimators suffer from high variance and an adverse dependence on the parameter dimensionality $d$, which has constrained their use to low-dimensional problems. In this work, we propose a policy-driven ZO framework that treats the sampling distribution over perturbation directions as a learnable policy and updates it to reduce the variance of directional estimates. We develop a practical algorithm implementing this idea and provide a theoretical analysis, showing that learned sampling distributions improve the quality of gradient information and relax the explicit dependence on $d$ in convergence bounds. Empirically, we validate the approach on challenging LLM fine-tuning benchmarks, demonstrating substantially improved performance compared to standard ZO baselines. Our results suggest that adaptive direction sampling is a promising route to make ZO fine-tuning viable at scale. The source code is available at https://github.com/brain-lab-research/zo_ldsd

Recent advances in Multimodal Large Language Models (MLLMs) have substantially driven the progress of autonomous agents for Graphical User Interface (GUI). Nevertheless, in real-world applications, GUI agents are often faced with non-stationary environments, leading to high computational costs for data curation and policy optimization. In this report, we introduce a novel MLLM-centered framework for GUI agents, which consists of two components: agentic-Q estimation and step-wise policy optimization. The former one aims to optimize a Q-model that can generate step-wise values to evaluate the contribution of a given action to task completion. The latter one takes step-wise samples from the state-action trajectory as inputs, and optimizes the policy via reinforcement learning with our agentic-Q model. It should be noticed that (i) all state-action trajectories are produced by the policy itself, so that the data collection costs are manageable; (ii) the policy update is decoupled from the environment, ensuring stable and efficient optimization. Empirical evaluations show that our framework endows Ovis2.5-9B with powerful GUI interaction capabilities, achieving remarkable performances on GUI navigation and grounding benchmarks and even surpassing contenders with larger scales.

In real-world federated learning (FL) systems, client participation is intermittent, heterogeneous, and often correlated with data characteristics or resource constraints. Existing fairness approaches in FL primarily focus on equalizing loss or accuracy conditional on participation, implicitly assuming that clients have comparable opportunities to contribute over time. However, when participation itself is uneven, these objectives can lead to systematic under-representation of intermittently available clients, even if per-round performance appears fair. We propose cumulative utility parity, a fairness principle that evaluates whether clients receive comparable long-term benefit per participation opportunity, rather than per training round. To operationalize this notion, we introduce availability-normalized cumulative utility, which disentangles unavoidable physical constraints from avoidable algorithmic bias arising from scheduling and aggregation. Experiments on temporally skewed, non-IID federated benchmarks demonstrate that our approach substantially improves long-term representation parity, while maintaining near-perfect performance.

We introduce KorMedMCQA-V, a Korean medical licensing-exam-style multimodal multiple-choice question answering benchmark for evaluating vision-language models (VLMs). The dataset consists of 1,534 questions with 2,043 associated images from Korean Medical Licensing Examinations (2012-2023), with about 30% containing multiple images requiring cross-image evidence integration. Images cover clinical modalities including X-ray, computed tomography (CT), electrocardiography (ECG), ultrasound, endoscopy, and other medical visuals. We benchmark over 50 VLMs across proprietary and open-source categories-spanning general-purpose, medical-specialized, and Korean-specialized families-under a unified zero-shot evaluation protocol. The best proprietary model (Gemini-3.0-Pro) achieves 96.9% accuracy, the best open-source model (Qwen3-VL-32B-Thinking) 83.7%, and the best Korean-specialized model (VARCO-VISION-2.0-14B) only 43.2%. We further find that reasoning-oriented model variants gain up to +20 percentage points over instruction-tuned counterparts, medical domain specialization yields inconsistent gains over strong general-purpose baselines, all models degrade on multi-image questions, and performance varies notably across imaging modalities. By complementing the text-only KorMedMCQA benchmark, KorMedMCQA-V forms a unified evaluation suite for Korean medical reasoning across text-only and multimodal conditions. The dataset is available via Hugging Face Datasets: https://huggingface.co/datasets/seongsubae/KorMedMCQA-V.

Time series chain (TSC) is a recently introduced concept that captures the evolving patterns in large scale time series. Informally, a time series chain is a temporally ordered set of subsequences, in which consecutive subsequences in the chain are similar to one another, but the last and the first subsequences maybe be dissimilar. Time series chain has the great potential to reveal latent unusual evolving trend in the time series, or identify precursor of important events in a complex system. Unfortunately, existing definitions of time series chains only consider finding chains in a single time series. As a result, they are likely to miss unexpected evolving patterns in interrupted time series, or across two related time series. To address this limitation, in this work, we introduce a new definition called \textit{Joint Time Series Chain}, which is specially designed for the task of finding unexpected evolving trend across interrupted time series or two related time series. Our definition focuses on mitigating the robustness issues caused by the gap or interruption in the time series. We further propose an effective ranking criterion to identify the best chain. We demonstrate that our proposed approach outperforms existing TSC work in locating unusual evolving patterns through extensive empirical evaluations. We further demonstrate the utility of our work with a real-life manufacturing application from Intel. Our source code is publicly available at the supporting page https://github.com/lizhang-ts/JointTSC .

Retrieval-augmented generation (RAG) is increasingly applied to question-answering over long academic papers, where accurate evidence allocation under a fixed token budget is critical. Existing approaches typically flatten academic papers into unstructured chunks during preprocessing, which destroys the native hierarchical structure. This loss forces retrieval to operate in a disordered space, thereby producing fragmented contexts, misallocating tokens to non-evidential regions under finite token budgets, and increasing the reasoning burden for downstream language models. To address these issues, we propose PT-RAG, an RAG framework that treats the native hierarchical structure of academic papers as a low-entropy retrieval prior. PT-RAG first inherits the native hierarchy to construct a structure-fidelity PaperTree index, which prevents entropy increase at the source. It then designs a path-guided retrieval mechanism that aligns query semantics to relevant sections and selects high relevance root-to-leaf paths under a fixed token budget, yielding compact, coherent, and low-entropy retrieval contexts. In contrast to existing RAG approaches, PT-RAG avoids entropy increase caused by destructive preprocessing and provides a native low-entropy structural basis for subsequent retrieval. To assess this design, we introduce entropy-based structural diagnostics that quantify retrieval fragmentation and evidence allocation accuracy. On three academic question-answering benchmarks, PT-RAG achieves consistently lower section entropy and evidence alignment cross entropy than strong baselines, indicating reduced context fragmentation and more precise allocation to evidential regions. These structural advantages directly translate into higher answer quality.

Existing robotic manipulation methods primarily rely on visual and proprioceptive observations, which may struggle to infer contact-related interaction states in partially observable real-world environments. Acoustic cues, by contrast, naturally encode rich interaction dynamics during contact, yet remain underexploited in current multimodal fusion literature. Most multimodal fusion approaches implicitly assume homogeneous roles across modalities, and thus design flat and symmetric fusion structures. However, this assumption is ill-suited for acoustic signals, which are inherently sparse and contact-driven. To achieve precise robotic manipulation through acoustic-informed perception, we propose a hierarchical representation fusion framework that progressively integrates audio, vision, and proprioception. Our approach first conditions visual and proprioceptive representations on acoustic cues, and then explicitly models higher-order cross-modal interactions to capture complementary dependencies among modalities. The fused representation is leveraged by a diffusion-based policy to directly generate continuous robot actions from multimodal observations. The combination of end-to-end learning and hierarchical fusion structure enables the policy to exploit task-relevant acoustic information while mitigating interference from less informative modalities. The proposed method has been evaluated on real-world robotic manipulation tasks, including liquid pouring and cabinet opening. Extensive experiment results demonstrate that our approach consistently outperforms state-of-the-art multimodal fusion frameworks, particularly in scenarios where acoustic cues provide task-relevant information not readily available from visual observations alone. Furthermore, a mutual information analysis is conducted to interpret the effect of audio cues in robotic manipulation via multimodal fusion.

With recent breakthroughs in large language models (LLMs) for reasoning, planning, and complex task generation, artificial intelligence systems are transitioning from isolated single-agent architectures to multi-agent systems with collaborative intelligence. However, in heterogeneous multi-agent systems (HMAS), capability differences among agents give rise to consistent cognitive problems, where strong and weak models fail to contribute effectively. We define the collaboration as a strong-weak system. Through comprehensive experiments, we disclose a counterintuitive phenomenon in the strong-weak system: a strong-weak collaboration may under-perform weak-weak combinations, revealing that cognitive mismatching are key bottlenecks limiting heterogeneous cooperation. To overcome these challenges, we propose an Entropy-Based Adaptive Guidance Framework that dynamically aligns the guidance with the cognitive state of each agent. The framework quantifies the understanding of weak agents through multi-dimensional entropy metrics - covering expression, uncertainty, structure, coherence, and relevance - and adaptively adjusts the intensity of the guidance at light, moderate and intensive levels. Furthermore, a Retrieval-Augmented Generation (RAG) mechanism is incorporated to retain successful collaboration experiences, enabling both immediate adaptation and long-term learning. Extensive experiments on three benchmark datasets, GSM8K, MBPP, and CVRP demonstrate that our approach consistently enhances the effectiveness and stability of heterogeneous collaboration. The results highlight that adaptive guidance not only mitigates cognitive imbalance but also establishes a scalable pathway toward more robust, cooperative multi-agent intelligence.

Neighborhood Aggregation Strategy (NAS) is a widely used approach in graph embedding, underpinning both Graph Neural Networks (GNNs) and Weisfeiler-Lehman (WL) methods. However, NAS-based methods are identified to be prone to over-smoothing-the loss of node distinguishability with increased iterations-thereby limiting their effectiveness. This paper identifies two characteristics in a network, i.e., the distributions of nodes and node degrees that are critical for expressive representation but have been overlooked in existing methods. We show that these overlooked characteristics contribute significantly to over-smoothing of NAS-methods. To address this, we propose a novel weighted distribution-aware kernel that embeds nodes while taking their distributional characteristics into consideration. Our method has three distinguishing features: (1) it is the first method to explicitly incorporate both distributional characteristics; (2) it requires no optimization; and (3) it effectively mitigates the adverse effects of over-smoothing, allowing WL to preserve node distinguishability and expressiveness even after many iterations of embedding. Experiments demonstrate that our method achieves superior community detection performance via spectral clustering, outperforming existing graph embedding methods, including deep learning methods, on standard benchmarks.

The expanding integration of Large Language Models (LLMs) into recommender systems poses critical challenges to evaluation reliability. This paper identifies and investigates a previously overlooked issue: benchmark data leakage in LLM-based recommendation. This phenomenon occurs when LLMs are exposed to and potentially memorize benchmark datasets during pre-training or fine-tuning, leading to artificially inflated performance metrics that fail to reflect true model performance. To validate this phenomenon, we simulate diverse data leakage scenarios by conducting continued pre-training of foundation models on strategically blended corpora, which include user-item interactions from both in-domain and out-of-domain sources. Our experiments reveal a dual-effect of data leakage: when the leaked data is domain-relevant, it induces substantial but spurious performance gains, misleadingly exaggerating the model's capability. In contrast, domain-irrelevant leakage typically degrades recommendation accuracy, highlighting the complex and contingent nature of this contamination. Our findings reveal that data leakage acts as a critical, previously unaccounted-for factor in LLM-based recommendation, which could impact the true model performance. We release our code at https://github.com/yusba1/LLMRec-Data-Leakage.

Anthropomorphic design is routinely used to make conversational agents more approachable and engaging. Yet its influence on users' perceptions remains poorly understood. Drawing on psychological theories, we propose that anthropomorphism influences risk perception via two complementary forms of trust, and that domain knowledge moderates these relationships. To test our model, we conducted a large-scale online experiment (N = 1,256) on a financial decision-support system implementing different anthropomorphic designs. We found that anthropomorphism indirectly reduces risk perception by increasing both cognitive and affective trust. Domain knowledge moderates these paths: participants with low financial knowledge experience a negative indirect effect of perceived anthropomorphism on risk perception via cognitive trust, whereas those with high financial knowledge exhibit a positive direct and indirect effect. We discuss theoretical contributions to human-AI interaction and design implications for calibrating trust in anthropomorphic decision-support systems for responsible AI.

We study parametric change-point detection, where the goal is to identify distributional changes in time series, under local differential privacy. In the non-private setting, we derive improved finite-sample accuracy guarantees for a change-point detection algorithm based on the generalized log-likelihood ratio test, via martingale methods. In the private setting, we propose two locally differentially private algorithms based on randomized response and binary mechanisms, and analyze their theoretical performance. We derive bounds on detection accuracy and validate our results through empirical evaluation. Our results characterize the statistical cost of local differential privacy in change-point detection and show how privacy degrades performance relative to a non-private benchmark. As part of this analysis, we establish a structural result for strong data processing inequalities (SDPI), proving that SDPI coefficients for Rényi divergences and their symmetric variants (Jeffreys-Rényi divergences) are achieved by binary input distributions. These results on SDPI coefficients are also of independent interest, with applications to statistical estimation, data compression, and Markov chain mixing.

We propose DiffusionRollout, a novel selective rollout planning strategy for autoregressive diffusion models, aimed at mitigating error accumulation in long-horizon predictions of physical systems governed by partial differential equations (PDEs). Building on the recently validated probabilistic approach to PDE solving, we further explore its ability to quantify predictive uncertainty and demonstrate a strong correlation between prediction errors and standard deviations computed over multiple samples-supporting their use as a proxy for the model's predictive confidence. Based on this observation, we introduce a mechanism that adaptively selects step sizes during autoregressive rollouts, improving long-term prediction reliability by reducing the compounding effect of conditioning on inaccurate prior outputs. Extensive evaluation on long-trajectory PDE prediction benchmarks validates the effectiveness of the proposed uncertainty measure and adaptive planning strategy, as evidenced by lower prediction errors and longer predicted trajectories that retain a high correlation with their ground truths.

Recently, large language models (LLMs) are extensively utilized to enhance development efficiency, leading to numerous benchmarks for evaluating their performance. However, these benchmarks predominantly focus on implementation, overlooking the equally critical aspect of software design. This gap raises two pivotal questions: (1) Can LLMs handle software design? (2) Can LLMs write code following the specific designs? To investigate these questions, this paper proposes DesBench, a design-aware benchmark for evaluating LLMs on three software design-related tasks: design-aware code generation, object-oriented modeling, and the design of acceptance test cases. DesBench comprises 30 manually crafted Java projects that include requirement documents, design models, implementations, and acceptance tests, amounting to a total of 30 design models, 194 Java classes, and 737 test cases. We evaluated seven state-of-the-art LLMs, including three DeepSeek R1, two Qwen2.5, and two GPT models, using DesBench. The results reveal that LLMs remain significantly challenged by the intricacies of software design: (1) For code generation, LLMs struggle to produce correct implementations when provided with only high-level or no designs. (2) In object-oriented modeling, while LLMs can accurately identify objects and classes, they face challenges in defining operations and inter-class relationships. (3) Acceptance test cases generated by LLMs from functional requirements achieve code coverage quality comparable to those written by humans. Our research highlights the current limitations of LLMs in managing software design and calls for further investigation into new design methodologies and languages suitable for LLM-based development.

Millimeter wave (mmWave) communication, utilizing beamforming techniques to address the inherent path loss limitation, is considered as one of the key technologies to support ever increasing high throughput and low latency demands of connected vehicles. However, adopting standard defined beamforming approach in highly dynamic vehicular environments often incurs high beam training overheads and reduction in the available airtime for communications, which is mainly due to exchanging pilot signals and exhaustive beam measurements. To this end, we present a multi-modal sensing and fusion learning framework as a potential alternative solution to reduce such overheads. In this framework, we first extract the representative features from the sensing modalities by modality specific encoders, then, utilize multi-head cross-modal attention to learn dependencies and correlations between different modalities, and subsequently fuse the multimodal features to obtain predicted top-k beams so that the best line-of-sight links can be proactively established. To show the generalizability of the proposed framework, we perform a comprehensive experiment in four different vehicle-to-infrastructure (V2I) and vehicle-to-vehicle (V2V) scenarios from real world multimodal and 60 GHz mmWave wireless sensing data. The experiment reveals that the proposed framework (i) achieves up to 96.72% accuracy on predicting top-15 beams correctly, (ii) incurs roughly 0.77 dB average power loss, and (iii) improves the overall latency and beam searching space overheads by 86.81% and 76.56% respectively for top-15 beams compared to standard defined approach.

We present \revise (\underline{Re}asoning with \underline{Vi}deo \underline{S}parsity), a multi-round agent for video question answering (VQA). Instead of uniformly sampling frames, \revise selects a small set of informative frames, maintains a summary-as-state across rounds, and stops early when confident. It supports proprietary vision-language models (VLMs) in a ``plug-and-play'' setting and enables reinforcement fine-tuning for open-source models. For fine-tuning, we introduce EAGER (Evidence-Adjusted Gain for Efficient Reasoning), an annotation-free reward with three terms: (1) Confidence gain: after new frames are added, we reward the increase in the log-odds gap between the correct option and the strongest alternative; (2) Summary sufficiency: at answer time we re-ask using only the last committed summary and reward success; (3) Correct-and-early stop: answering correctly within a small turn budget is rewarded. Across multiple VQA benchmarks, \revise improves accuracy while reducing frames, rounds, and prompt tokens, demonstrating practical sparse video reasoning.

Neural scaling laws provide a predictable recipe for AI advancement: reducing numerical precision should linearly improve computational efficiency and energy profile (E proportional to bits). In this paper, we demonstrate that this scaling law breaks in the context of multi-hop reasoning. We reveal a 'quantization trap' where reducing precision from 16-bit to 8/4-bit paradoxically increases more net energy consumption while degrading reasoning accuracy. We provide a rigorous theoretical decomposition that attributes this failure to hardware casting overhead, the hidden latency cost of dequantization kernels, which becomes a dominant bottleneck in sequential reasoning chains, as well as to a sequential energy amortization failure. As a result, scaling law breaking is unavoidable in practice. Our findings suggest that the industry's "smaller-is-better" heuristic is mathematically counterproductive for complex reasoning tasks.

Agentic AI require persistent memory to store user-specific histories beyond the limited context window of LLMs. Existing memory systems use dense vector databases or knowledge-graph traversal (or hybrid), incurring high retrieval latency and poor storage scalability. We introduce Hippocampus, an agentic memory management system that uses compact binary signatures for semantic search and lossless token-ID streams for exact content reconstruction. Its core is a Dynamic Wavelet Matrix (DWM) that compresses and co-indexes both streams to support ultra-fast search in the compressed domain, thus avoiding costly dense-vector or graph computations. This design scales linearly with memory size, making it suitable for long-horizon agentic deployments. Empirically, our evaluation shows that Hippocampus reduces end-to-end retrieval latency by up to 31$\times$ and cuts per-query token footprint by up to 14$\times$, while maintaining accuracy on both LoCoMo and LongMemEval benchmarks.

Inspired by the human visual system, which operates on two parallel yet interactive streams for contextual and spatial understanding, this article presents Two Interactive Streams (TwInS), a novel bio-inspired joint learning framework capable of simultaneously performing scene parsing and geometric vision tasks. TwInS adopts a unified, general-purpose architecture in which multi-level contextual features from the scene parsing stream are infused into the geometric vision stream to guide its iterative refinement. In the reverse direction, decoded geometric features are projected into the contextual feature space for selective heterogeneous feature fusion via a novel cross-task adapter, which leverages rich cross-view geometric cues to enhance scene parsing. To eliminate the dependence on costly human-annotated correspondence ground truth, TwInS is further equipped with a tailored semi-supervised training strategy, which unleashes the potential of large-scale multi-view data and enables continuous self-evolution without requiring ground-truth correspondences. Extensive experiments conducted on three public datasets validate the effectiveness of TwInS's core components and demonstrate its superior performance over existing state-of-the-art approaches. The source code will be made publicly available upon publication.

This monograph reports a multi-agent proof sprint on ten research-level problems, combining rapid draft generation with adversarial verification, targeted repair, and explicit provenance. The workflow uses wiring-diagram decompositions of claim dependencies to localize gaps and coordinate reviewer-driven revisions. Final outcomes are heterogeneous but explicit: the manuscript distinguishes mathematical status from QC-validation status. Mathematically, Problem~3 has a validation-complete existence path under the scoped criterion used here (uniqueness/irreducibility treated as optional), Problem 5 is solved in a scope-limited form for $F_O$-local connective spectra, Problem 10 is conditional under clearly stated assumptions (with explicit necessity counterexamples when assumptions are dropped), and Problems 4 and 6 are partial with named remaining obligations in the general case (including an unconditional $K_n$ result for Problem 6 with $c_0 = 1/3$). Problem 7 is treated as provisionally closed via the rotation-route theorem chain, pending independent ledger re-check. At the QC layer, Problems~7 and~9 have node-level validation artifacts but still contain unresolved verifier gaps. The main methodological result is that structure-aware verification and layer-switching strategies improve reliability and calibration in compressed proof sprints.

Clustering serves as a vital tool for uncovering latent data structures, and achieving both high accuracy and interpretability is essential. To this end, existing methods typically construct binary decision trees by solving mixed-integer nonlinear optimization problems, often leading to significant computational costs and suboptimal solutions. Furthermore, binary decision trees frequently result in excessively deep structures, which makes them difficult to interpret. To mitigate these issues, we propose an interpretable clustering method based on optimal multiway-split decision trees, formulated as a 0-1 integer linear optimization problem. This reformulation renders the optimization problem more tractable compared to existing models. A key feature of our method is the integration of a one-dimensional K-means algorithm for the discretization of continuous variables, allowing for flexible and data-driven branching. Extensive numerical experiments on publicly available real-world datasets demonstrate that our method outperforms baseline methods in terms of clustering accuracy and interpretability. Our method yields multiway-split decision trees with concise decision rules while maintaining competitive performance across various evaluation metrics.

Rule learning-based models are widely used in highly interpretable scenarios due to their transparent structures. Inductive logic programming (ILP), a form of machine learning, induces rules from facts while maintaining interpretability. Differentiable ILP models enhance this process by leveraging neural networks to improve robustness and scalability. However, most differentiable ILP methods rely on symbolic datasets, facing challenges when learning directly from raw data. Specifically, they struggle with explicit label leakage: The inability to map continuous inputs to symbolic variables without explicit supervision of input feature labels. In this work, we address this issue by integrating a self-supervised differentiable clustering model with a novel differentiable ILP model, enabling rule learning from raw data without explicit label leakage. The learned rules effectively describe raw data through its features. We demonstrate that our method intuitively and precisely learns generalized rules from time series and image data.

Evaluation and alignment pipelines for large language models increasingly rely on LLM-based judges, whose behavior is guided by natural-language rubrics and validated on benchmarks. We identify a previously under-recognized vulnerability in this workflow, which we term Rubric-Induced Preference Drift (RIPD). Even when rubric edits pass benchmark validation, they can still produce systematic and directional shifts in a judge's preferences on target domains. Because rubrics serve as a high-level decision interface, such drift can emerge from seemingly natural, criterion-preserving edits and remain difficult to detect through aggregate benchmark metrics or limited spot-checking. We further show this vulnerability can be exploited through rubric-based preference attacks, in which benchmark-compliant rubric edits steer judgments away from a fixed human or trusted reference on target domains, systematically inducing RIPD and reducing target-domain accuracy up to 9.5% (helpfulness) and 27.9% (harmlessness). When these judgments are used to generate preference labels for downstream post-training, the induced bias propagates through alignment pipelines and becomes internalized in trained policies. This leads to persistent and systematic drift in model behavior. Overall, our findings highlight evaluation rubrics as a sensitive and manipulable control interface, revealing a system-level alignment risk that extends beyond evaluator reliability alone. The code is available at: https://github.com/ZDCSlab/Rubrics-as-an-Attack-Surface. Warning: Certain sections may contain potentially harmful content that may not be appropriate for all readers.

Current alignment methods for Large Language Models (LLMs) rely on compressing vast amounts of human preference data into static, absolute reward functions, leading to data scarcity, noise sensitivity, and training instability. We introduce Elo-Evolve, a co-evolutionary framework that redefines alignment as dynamic multi-agent competition within an adaptive opponent pool. Our approach makes two key innovations: (1) eliminating Bradley-Terry model dependencies by learning directly from binary win/loss outcomes in pairwise competitions, and (2) implementing Elo-orchestrated opponent selection that provides automatic curriculum learning through temperature-controlled sampling. We ground our approach in PAC learning theory, demonstrating that pairwise comparison achieves superior sample complexity and empirically validate a 4.5x noise reduction compared to absolute scoring approaches. Experimentally, we train a Qwen2.5-7B model using our framework with opponents including Qwen2.5-14B, Qwen2.5-32B, and Qwen3-8B models. Results demonstrate a clear performance hierarchy: point-based methods < static pairwise training < Elo-Evolve across Alpaca Eval 2.0 and MT-Bench, validating the progressive benefits of pairwise comparison and dynamic opponent selection for LLM alignment.

Proof-of-Vulnerability (PoV) generation is a critical task in software security, serving as a cornerstone for vulnerability validation, false positive reduction, and patch verification. While directed fuzzing effectively drives path exploration, satisfying complex semantic constraints remains a persistent bottleneck in automated exploit generation. Large Language Models (LLMs) offer a promising alternative with their semantic reasoning capabilities; however, existing LLM-based approaches lack sufficient grounding in concrete execution behavior, limiting their ability to generate precise PoVs. In this paper, we present DrillAgent, an agentic framework that reformulates PoV generation as an iterative hypothesis-verification-refinement process. To bridge the gap between static reasoning and dynamic execution, DrillAgent synergizes LLM-based semantic inference with feedback from concrete program states. The agent analyzes the target code to hypothesize inputs, observes execution behavior, and employs a novel mechanism to translate low-level execution traces into source-level constraints. This closed-loop design enables the agent to incrementally align its input generation with the precise requirements of the vulnerability. We evaluate DrillAgent on SEC-bench, a large-scale benchmark of real-world C/C++ vulnerabilities. Experimental results show that DrillAgent substantially outperforms state-of-the-art LLM agent baselines under fixed budget constraints, solving up to 52.8% more CVE tasks than the best-performing baseline. These results highlight the necessity of execution-state-aware reasoning for reliable PoV generation in complex software systems.

Large language models (LLMs) have revolutionized natural language processing, yet hallucinations in knowledge-intensive tasks remain a critical challenge. Retrieval-augmented generation (RAG) addresses this by integrating external knowledge, but its efficacy depends on accurate document retrieval and ranking. Although existing rerankers demonstrate effectiveness, they frequently necessitate specialized training, impose substantial computational expenses, and fail to fully exploit the semantic capabilities of LLMs, particularly their inherent confidence signals. We propose the LLM-Confidence Reranker (LCR), a training-free, plug-and-play algorithm that enhances reranking in RAG systems by leveraging black-box LLM confidence derived from Maximum Semantic Cluster Proportion (MSCP). LCR employs a two-stage process: confidence assessment via multinomial sampling and clustering, followed by binning and multi-level sorting based on query and document confidence thresholds. This approach prioritizes relevant documents while preserving original rankings for high-confidence queries, ensuring robustness. Evaluated on BEIR and TREC benchmarks with BM25 and Contriever retrievers, LCR--using only 7--9B-parameter pre-trained LLMs--consistently improves NDCG@5 by up to 20.6% across pre-trained LLM and fine-tuned Transformer rerankers, without degradation. Ablation studies validate the hypothesis that LLM confidence positively correlates with document relevance, elucidating LCR's mechanism. LCR offers computational efficiency, parallelism for scalability, and broad compatibility, mitigating hallucinations in applications like medical diagnosis.

Large language models (LLMs) increasingly operate in environments where they encounter social information such as other agents' answers, tool outputs, or human recommendations. In humans, such inputs influence judgments in ways that depend on the source's credibility and the strength of consensus. This paper investigates whether LLMs exhibit analogous patterns of influence and whether they privilege feedback from humans over feedback from other LLMs. Across three binary decision-making tasks, reading comprehension, multi-step reasoning, and moral judgment, we present four instruction-tuned LLMs with prior responses attributed either to friends, to human experts, or to other LLMs. We manipulate whether the group is correct and vary the group size. In a second experiment, we introduce direct disagreement between a single human and a single LLM. Across tasks, models conform significantly more to responses labeled as coming from human experts, including when that signal is incorrect, and revise their answers toward experts more readily than toward other LLMs. These results reveal that expert framing acts as a strong prior for contemporary LLMs, suggesting a form of credibility-sensitive social influence that generalizes across decision domains.

Standard Knowledge Distillation (KD) compresses Large Language Models (LLMs) by optimizing final outputs, yet it typically treats the teacher's intermediate layer's thought process as a black box. While feature-based distillation attempts to bridge this gap, existing methods (e.g., MSE and asymmetric KL divergence) ignore the rich uncertainty profiles required for the final output. In this paper, we introduce DistillLens, a framework that symmetrically aligns the evolving thought processes of student and teacher models. By projecting intermediate hidden states into the vocabulary space via the Logit Lens, we enforce structural alignment using a symmetric divergence objective. Our analysis proves that this constraint imposes a dual-sided penalty, preventing both overconfidence and underconfidence while preserving the high-entropy information conduits essential for final deduction. Extensive experiments on GPT-2 and Llama architectures demonstrate that DistillLens consistently outperforms standard KD and feature-transfer baselines on diverse instruction-following benchmarks. The code is available at https://github.com/manishdhakal/DistillLens.

We address multi-objective unmanned aerial vehicle (UAV) placement for motorway intelligent transportation systems, where deployments must balance coverage, link quality, and UAV count under geometric constraints. We construct a reproducible benchmark from highD motorway recordings with recording-level splits and generate Pareto-optimal labels via NSGA-II. A preference rule yields deployable targets while preserving multi-objective evaluation. We train fast surrogate models that map unordered vehicle positions to UAV count and continuous placements, using permutation-aware losses and constraint-regularized training across set-based and sequence-based architectures. The evaluation protocol combines Pareto quality metrics, success-rate curves, runtime benchmarks, and robustness studies, with uncertainty quantified by recording-level bootstrap. Results indicate that permutation-invariant set models provide the strongest coverage--SNR--count trade-off among learned predictors and approach NSGA-II quality while enabling real-time inference. Under shared budgets, they offer a more favorable success--latency trade-off than heuristic baselines. The benchmark, splits are released to support reproducible ITS deployment studies and to facilitate comparisons under shared operational budgets.

While reasoning models have achieved remarkable success in complex reasoning tasks, their increasing power necessitates stringent safety measures. For safety alignment, the core challenge lies in the inherent trade-off between safety and utility. However, prevailing alignment strategies typically construct CoT training data with explicit safety rules via context distillation. This approach inadvertently limits reasoning capabilities by creating a rigid association between rule memorization and refusal. To mitigate the safety-utility trade-off, we propose the Adaptive Safe Context Learning (ASCL) framework to improve the reasoning given proper context. ASCL formulates safety alignment as a multi-turn tool-use process, empowering the model to autonomously decide when to consult safety rules and how to generate the ongoing reasoning. Furthermore, to counteract the preference for rule consultation during RL, we introduce Inverse Frequency Policy Optimization (IFPO) to rebalance advantage estimates. By decoupling rule retrieval and subsequent reasoning, our method achieves higher overall performance compared to baselines.

To fulfill user instructions, autonomous web agents must contend with the inherent complexity and volatile nature of real-world websites. Conventional paradigms predominantly rely on Supervised Fine-Tuning (SFT) or Offline Reinforcement Learning (RL) using static datasets. However, these methods suffer from severe distributional shifts, as offline trajectories fail to capture the stochastic state transitions and real-time feedback of unconstrained wide web environments. In this paper, we propose a robust Online Reinforcement Learning WebAgent, designed to optimize its policy through direct, iterative interactions with unconstrained wide websites. Our approach comprises three core innovations: 1) Hierarchical Multi-Task Fine-tuning: We curate a comprehensive mixture of datasets categorized by functional primitives -- Planning, Acting, and Grounding -- establishing a Vision-Language Model (VLM) with strong instruction-following capabilities for Web GUI tasks. 2) Online Agentic RL in the Wild: We develop an online interaction environment and fine-tune the VLM using a specialized RL pipeline. We introduce a Hybrid Reward Mechanism that combines a ground-truth-agnostic WebJudge for holistic outcome assessment with a Rule-based Decision Tree (RDT) for progress reward. This system effectively mitigates the credit assignment challenge in long-horizon navigation. Notably, our RL-enhanced model achieves a 38.1\% success rate (pass@5) on WebArena, outperforming all existing monolithic baselines. 3) Operator Agent: We introduce a modular agentic framework, namely \textbf{OpAgent}, orchestrating a Planner, Grounder, Reflector, and Summarizer. This synergy enables robust error recovery and self-correction, elevating the agent's performance to a new State-of-the-Art (SOTA) success rate of \textbf{71.6\%}.

Semantic Communication (SemCom) has emerged as a promising paradigm for 6G networks, aiming to extract and transmit task-relevant information rather than minimizing bit errors. However, applying SemCom to realistic downlink Multi-User Multi-Input Multi-Output (MU-MIMO) Orthogonal Frequency Division Multiplexing (OFDM) systems remains challenging due to severe Multi-User Interference (MUI) and frequency-selective fading. Existing Deep Joint Source-Channel Coding (DJSCC) schemes, primarily designed for point-to-point links, suffer from performance saturation in multi-user scenarios. To address these issues, we propose a scenario-adaptive MU-MIMO SemCom framework featuring an asymmetric architecture tailored for downlink transmission. At the transmitter, we introduce a scenario-aware semantic encoder that dynamically adjusts feature extraction based on Channel State Information (CSI) and Signal-to-Noise Ratio (SNR), followed by a neural precoding network designed to mitigate MUI in the semantic domain. At the receiver, a lightweight decoder equipped with a novel pilot-guided attention mechanism is employed to implicitly perform channel equalization and feature calibration using reference pilot symbols. Extensive simulation results over 3GPP channel models demonstrate that the proposed framework significantly outperforms DJSCC and traditional Separate Source-Channel Coding (SSCC) schemes in terms of Peak Signal-to-Noise Ratio (PSNR) and classification accuracy, particularly in low-SNR regimes, while maintaining low latency and computational cost on edge devices.

We introduce Discernment, a semantic communication system that transmits the meaning of physical signals (baseband radio and audio) over a technical channel using GenAI models operating in discrete spaces. Discernment dynamically adapts to channel impairments - modeled as erasure channels - by switching between an autoregressive or a diffusion-based generative algorithm, depending on the erasure pattern. Our results show that Discernment maintains semantic integrity even as channel capacity severely degrades, exhibiting very small and graceful performance decline in both classification accuracy and statistical fidelity of the reconstructed meaning. These findings demonstrate Discernment's ability to adjust to diverse physical channel conditions while maintaining spectral efficiency and low model complexity, making it well suited for IoT deployments and strongly motivating further research on this semantic channel paradigm.

Cooperative perception systems for autonomous driving aim to overcome the limited perception range of a single vehicle by communicating with adjacent agents to share sensing information. While this improves perception performance, these systems also face a significant privacy-leakage issue, as sensitive visual content can potentially be reconstructed from the shared data. In this paper, we propose a novel Privacy-Concealing Cooperation (PCC) framework for Bird's Eye View (BEV) semantic segmentation. Based on commonly shared BEV features, we design a hiding network to prevent an image reconstruction network from recovering the input images from the shared features. An adversarial learning mechanism is employed to train the network, where the hiding network works to conceal the visual clues in the BEV features while the reconstruction network attempts to uncover these clues. To maintain segmentation performance, the perception network is integrated with the hiding network and optimized end-to-end. The experimental results demonstrate that the proposed PCC framework effectively degrades the quality of the reconstructed images with minimal impact on segmentation performance, providing privacy protection for cooperating vehicles. The source code will be made publicly available upon publication.

Reward models (RMs) play a central role throughout the language model (LM) pipeline, particularly in non-verifiable domains. However, the dominant LLM-as-a-Judge paradigm relies on the strong reasoning capabilities of large models, while alternative approaches require reference responses or explicit rubrics, limiting flexibility and broader accessibility. In this work, we propose FLIP (FLipped Inference for Prompt reconstruction), a reference-free and rubric-free reward modeling approach that reformulates reward modeling through backward inference: inferring the instruction that would most plausibly produce a given response. The similarity between the inferred and the original instructions is then used as the reward signal. Evaluations across four domains using 13 small language models show that FLIP outperforms LLM-as-a-Judge baselines by an average of 79.6%. Moreover, FLIP substantially improves downstream performance in extrinsic evaluations under test-time scaling via parallel sampling and GRPO training. We further find that FLIP is particularly effective for longer outputs and robust to common forms of reward hacking. By explicitly exploiting the validation-generation gap, FLIP enables reliable reward modeling in downscaled regimes where judgment methods fail. Code available at https://github.com/yikee/FLIP.

As breakthroughs in deep learning transform key industries, models are increasingly required to extrapolate on datapoints found outside the range of the training set, a challenge we coin as out-of-support (OoS) generalisation. However, neural networks frequently exhibit catastrophic failure on OoS samples, yielding unrealistic but overconfident predictions. We address this challenge by reformulating the OoS generalisation problem as a sequence modelling task in the weight space, wherein the training set is partitioned into concentric shells corresponding to discrete sequential steps. Our WeightCaster framework yields plausible, interpretable, and uncertainty-aware predictions without necessitating explicit inductive biases, all the while maintaining high computational efficiency. Emprical validation on a synthetic cosine dataset and real-world air quality sensor readings demonstrates performance competitive or superior to the state-of-the-art. By enhancing reliability beyond in-distribution scenarios, these results hold significant implications for the wider adoption of artificial intelligence in safety-critical applications.

Large language models (LLMs) remain vulnerable to jailbreak prompts that elicit harmful or policy-violating outputs, while many existing defenses rely on expensive fine-tuning, intrusive prompt rewriting, or external guardrails that add latency and can degrade helpfulness. We present AISA, a lightweight, single-pass defense that activates safety behaviors already latent inside the model rather than treating safety as an add-on. AISA first localizes intrinsic safety awareness via spatiotemporal analysis and shows that intent-discriminative signals are broadly encoded, with especially strong separability appearing in the scaled dot-product outputs of specific attention heads near the final structural tokens before generation. Using a compact set of automatically selected heads, AISA extracts an interpretable prompt-risk score with minimal overhead, achieving detector-level performance competitive with strong proprietary baselines on small (7B) models. AISA then performs logits-level steering: it modulates the decoding distribution in proportion to the inferred risk, ranging from normal generation for benign prompts to calibrated refusal for high-risk requests -- without changing model parameters, adding auxiliary modules, or requiring multi-pass inference. Extensive experiments spanning 13 datasets, 12 LLMs, and 14 baselines demonstrate that AISA improves robustness and transfer while preserving utility and reducing false refusals, enabling safer deployment even for weakly aligned or intentionally risky model variants.

Large Language Models (LLMs) with agentic web search capabilities show strong potential for tasks requiring real-time information access and complex fact retrieval, yet evaluating such systems remains challenging. We introduce \bench, a rigorous and regularly updated benchmark designed to assess the agentic web search abilities of LLMs. \bench automatically generates fresh question-answer pairs from recent news articles, ensuring that questions require information beyond an LLM's training data and enabling clear separation between internal knowledge and search capability. The benchmark features intentionally difficult questions requiring multi-hop search queries, page visits, and reasoning, making it well-suited for evaluating agentic search behavior. Our automated data curation and question generation pipeline enables frequent benchmark updates and supports construction of a large-scale training dataset for agentic web search models, addressing the scarcity of such data in the research community. To ensure reliable evaluation, we include a subset of human-verified samples in the test set. We evaluate a broad range of systems using \bench, including commercial and open-weight LLMs as well as LLM-based web search APIs. The leaderboard, datasets, and code are publicly available at livenewsbench.com.

Small Island Developing States (SIDS) are disproportionately exposed to climate-driven disasters, yet often rely on fragile terrestrial networks that fail when they are most needed. TV White Space (TVWS) links offer long-range, low-power coverage; however, current deployments depend on Protocol to Access White Spaces (PAWS) database connectivity for channel authorization, creating a single point of failure during outages. We present SIDSense, an edge AI framework for database-free TVWS operation that preserves regulatory intent through a compliance-gated controller, audit logging, and graceful degradation. SIDSense couples CNN-based spectrum classification with a hybrid sensing-first, authorization-as-soon-as-possible workflow and co-locates sensing and video enhancement with a private 5G stack on a maritime vessel to sustain situational-awareness video backhaul. Field experiments in Barbados demonstrate sustained connectivity during simulated PAWS outages, achieving 94.2% sensing accuracy over 470-698 MHz with 23 ms mean decision latency, while maintaining zero missed 5G Layer-1 (L1) deadlines under GPU-aware scheduling. We release an empirical Caribbean TVWS propagation and occupancy dataset and look to contribute some of the components of the SIDSense pipeline to the open source community to accelerate resilient connectivity deployments in climate-vulnerable regions.

Large language models (LLMs) are widely deployed as general-purpose problem solvers, making accurate confidence estimation critical for reliable use. Prior work on LLM calibration largely focuses on response-level confidence, which estimates the correctness of a single generated output. However, this formulation is misaligned with many practical settings where the central question is how likely a model is to solve a query overall. We show that this mismatch results from the stochastic nature of modern LLM decoding, under which single-response correctness fails to reflect underlying model capability. To address this issue, we introduce capability calibration, which targets the model's expected accuracy on a query. We formally distinguish capability calibration from response calibration and show that the two differ both theoretically and empirically. We establish an empirical evaluation setup and study a range of confidence estimation methods. Our results demonstrate that capability-calibrated confidence improves pass@$k$ prediction and inference budget allocation, establishing a foundation with potential for diverse applications.

In this paper, we propose a fast algorithm for element selection, a multiplication-free form of dimension reduction that produces a dimension-reduced vector by simply selecting a subset of elements from the input. Dimension reduction is a fundamental technique for reducing unnecessary model parameters, mitigating overfitting, and accelerating training and inference. A standard approach is principal component analysis (PCA), but PCA relies on matrix multiplications; on resource-constrained systems, the multiplication count itself can become a bottleneck. Element selection eliminates this cost because the reduction consists only of selecting elements, and thus the key challenge is to determine which elements should be retained. We evaluate a candidate subset through the minimum mean-squared error of linear regression that predicts a target vector from the selected elements, where the target may be, for example, a one-hot label vector in classification. When an explicit target is unavailable, the input itself can be used as the target, yielding a reconstruction-based criterion. The resulting optimization is combinatorial, and exhaustive search is impractical. To address this, we derive an efficient formula for the objective change caused by swapping a selected and an unselected element, using the matrix inversion lemma, and we perform a swap-based local search that repeatedly applies objective-decreasing swaps until no further improvement is possible. Experiments on MNIST handwritten-digit images demonstrate the effectiveness of the proposed method.

Quantum reservoir computing offers a promising route for time series learning by modelling sequential data via rich quantum dynamics while the only training required happens at the level of a lightweight classical readout. However, studies featuring efficient and implementable quantum reservoir architectures along with model learning guarantees remain scarce in the literature. To close this gap, we introduce QuaRK, an end-to-end framework that couples a hardware-realistic quantum reservoir featurizer with a kernel-based readout scheme. Given a sequence of sample points, the reservoir injects the points one after the other to yield a compact feature vector from efficiently measured k-local observables using classical shadow tomography, after which a classical kernel-based readout learns the target mapping with explicit regularization and fast optimization. The resulting pipeline exposes clear computational knobs -- circuit width and depth as well as the measurement budget -- while preserving the flexibility of kernel methods to model nonlinear temporal functionals and being scalable to high-dimensional data. We further provide learning-theoretic generalization guarantees for dependent temporal data, linking design and resource choices to finite-sample performance, thereby offering principled guidance for building reliable temporal learners. Empirical experiments validate QuaRK and illustrate the predicted interpolation and generalization behaviours on synthetic beta-mixing time series tasks.

Humans excel at remembering concrete experiences along spatiotemporal contexts and performing reasoning across those events, i.e., the capacity for episodic memory. In contrast, memory in language agents remains mainly semantic, and current agents are not yet capable of effectively recollecting and reasoning over interaction histories. We identify and formalize the core challenges of episodic recollection and reasoning from this gap, and observe that existing work often overlooks episodicity, lacks explicit event modeling, or overemphasizes simple retrieval rather than complex reasoning. We present REMem, a two-phase framework for constructing and reasoning with episodic memory: 1) Offline indexing, where REMem converts experiences into a hybrid memory graph that flexibly links time-aware gists and facts. 2) Online inference, where REMem employs an agentic retriever with carefully curated tools for iterative retrieval over the memory graph. Comprehensive evaluation across four episodic memory benchmarks shows that REMem substantially outperforms state-of-the-art memory systems such as Mem0 and HippoRAG 2, showing 3.4% and 13.4% absolute improvements on episodic recollection and reasoning tasks, respectively. Moreover, REMem also demonstrates more robust refusal behavior for unanswerable questions.

Federated learning (FL) enables collaborative training across organizational silos without sharing raw data, making it attractive for privacy-sensitive applications. With the rapid adoption of large language models (LLMs), federated fine-tuning of generative LLMs has gained attention as a way to leverage distributed data while preserving confidentiality. However, this setting introduces fundamental challenges: (i) privacy leakage of personally identifiable information (PII) due to LLM memorization, and (ii) a persistent tension between global generalization and local utility under heterogeneous data. Existing defenses, such as data sanitization and differential privacy, reduce leakage but often degrade downstream performance. We propose SecureGate, a privacy-aware federated fine-tuning framework for LLMs that provides fine-grained privacy control without sacrificing utility. SecureGate employs a dual-adapter LoRA architecture: a secure adapter that learns sanitized, globally shareable representations, and a revealing adapter that captures sensitive, organization-specific knowledge. A token-controlled gating module selectively activates these adapters at inference time, enabling controlled information disclosure without retraining. Extensive experiments across multiple LLMs and real-world datasets show that SecureGate improves task utility while substantially reducing PII leakage, achieving up to a 31.66X reduction in inference attack accuracy and a 17.07X reduction in extraction recall for unauthorized requests. Additionally, it maintains 100% routing reliability to the correct adapter and incurs only minimal computational and communication overhead.

Identifying feature representations in language models is a central task in mechanistic interpretability. Several recent studies have made an implicit assumption that feature representations can be inferred in some cases from singular vectors of attention matrices. However, sound justification for this assumption is lacking. In this paper we address that question, asking: why and when do singular vectors align with features? First, we demonstrate that singular vectors robustly align with features in a model where features can be directly observed. We then show theoretically that such alignment is expected under a range of conditions. We close by asking how, operationally, alignment may be recognized in real models where feature representations are not directly observable. We identify sparse attention decomposition as a testable prediction of alignment, and show evidence that it emerges in a manner consistent with predictions in real models. Together these results suggest that alignment of singular vectors with features can be a sound and theoretically justified basis for feature identification in language models.

Natural language to SQL (NL2SQL) translation enables non-expert users to query relational databases through natural language. Recently, NL2SQL agents, powered by the reasoning capabilities of Large Language Models (LLMs), have significantly advanced NL2SQL translation. Nonetheless, NL2SQL agents still make mistakes when faced with large-scale real-world databases because they lack knowledge of how to correctly leverage the underlying data (e.g., knowledge about the intent of each column) and form misconceptions about the data when querying it, leading to errors. Prior work has studied generating facts about the database to provide more context to NL2SQL agents, but such approaches simply restate database contents without addressing the agent's misconceptions. In this paper, we propose Tk-Boost, a bolt-on framework for augmenting any NL2SQL agent with tribal knowledge: knowledge that corrects the agent's misconceptions in querying the database accumulated through experience using the database. To accumulate experience, Tk-Boost first asks the NL2SQL agent to answer a few queries on the database, identifies the agent's misconceptions by analyzing its mistakes on the database, and generates tribal knowledge to address them. To enable accurate retrieval, Tk-Boost indexes this knowledge with applicability conditions that specify the query features for which the knowledge is useful. When answering new queries, Tk-Boost uses this knowledge to provide feedback to the NL2SQL agent, resolving the agent's misconceptions during SQL generation, and thus improving the agent's accuracy. Extensive experiments across the BIRD and Spider 2.0 benchmarks with various NL2SQL agents shows Tk-Boost improves NL2SQL agents accuracy by up to 16.9% on Spider 2.0 and 13.7% on BIRD

Large language models (LLMs) have demonstrated impressive reasoning capabilities by scaling test-time compute via long Chain-of-Thought (CoT). However, recent findings suggest that raw token counts are unreliable proxies for reasoning quality: increased generation length does not consistently correlate with accuracy and may instead signal "overthinking," leading to performance degradation. In this work, we quantify inference-time effort by identifying deep-thinking tokens -- tokens where internal predictions undergo significant revisions in deeper model layers prior to convergence. Across four challenging mathematical and scientific benchmarks (AIME 24/25, HMMT 25, and GPQA-diamond) and a diverse set of reasoning-focused models (GPT-OSS, DeepSeek-R1, and Qwen3), we show that deep-thinking ratio (the proportion of deep-thinking tokens in a generated sequence) exhibits a robust and consistently positive correlation with accuracy, substantially outperforming both length-based and confidence-based baselines. Leveraging this insight, we introduce Think@n, a test-time scaling strategy that prioritizes samples with high deep-thinking ratios. We demonstrate that Think@n matches or exceeds standard self-consistency performance while significantly reducing inference costs by enabling the early rejection of unpromising generations based on short prefixes.

LLM-powered agents are beginning to automate user's tasks across the open web, often with access to user resources such as emails and calendars. Unlike standard LLMs answering questions in a controlled ChatBot setting, web agents act "in the wild", interacting with third parties and leaving behind an action trace. Therefore, we ask the question: how do web agents handle user resources when accomplishing tasks on their behalf across live websites? In this paper, we formalize Natural Agentic Oversharing -- the unintentional disclosure of task-irrelevant user information through an agent trace of actions on the web. We introduce SPILLage, a framework that characterizes oversharing along two dimensions: channel (content vs. behavior) and directness (explicit vs. implicit). This taxonomy reveals a critical blind spot: while prior work focuses on text leakage, web agents also overshare behaviorally through clicks, scrolls, and navigation patterns that can be monitored. We benchmark 180 tasks on live e-commerce sites with ground-truth annotations separating task-relevant from task-irrelevant attributes. Across 1,080 runs spanning two agentic frameworks and three backbone LLMs, we demonstrate that oversharing is pervasive with behavioral oversharing dominates content oversharing by 5x. This effect persists -- and can even worsen -- under prompt-level mitigation. However, removing task-irrelevant information before execution improves task success by up to 17.9%, demonstrating that reducing oversharing improves task success. Our findings underscore that protecting privacy in web agents is a fundamental challenge, requiring a broader view of "output" that accounts for what agents do on the web, not just what they type. Our datasets and code are available at https://github.com/jrohsc/SPILLage.

Many training-free sparse attention methods are effective for accelerating diffusion models. Recently, several works suggest that making sparse attention trainable can further increase sparsity while preserving generation quality. We study three key questions: (1) when do the two common masking rules, i.e., Top-k and Top-p, fail, and how can we avoid these failures? (2) why can trainable sparse attention reach higher sparsity than training-free methods? (3) what are the limitations of fine-tuning sparse attention using the diffusion loss, and how can we address them? Based on this analysis, we propose SpargeAttention2, a trainable sparse attention method that achieves high sparsity without degrading generation quality. SpargeAttention2 includes (i) a hybrid masking rule that combines Top-k and Top-p for more robust masking at high sparsity, (ii) an efficient trainable sparse attention implementation, and (iii) a distillation-inspired fine-tuning objective to better preserve generation quality during fine-tuning using sparse attention. Experiments on video diffusion models show that SpargeAttention2 reaches 95% attention sparsity and a 16.2x attention speedup while maintaining generation quality, consistently outperforming prior sparse attention methods.

In this work, we investigate the use of data-driven equation discovery for dynamical systems to model and forecast continuous-time dynamics of unconstrained optimization problems. To avoid expensive evaluations of the objective function and its gradient, we leverage trajectory data on the optimization variables to learn the continuous-time dynamics associated with gradient descent, Newton's method, and ADAM optimization. The discovered gradient flows are then solved as a surrogate for the original optimization problem. To this end, we introduce the Learned Gradient Flow (LGF) optimizer, which is equipped to build surrogate models of variable polynomial order in full- or reduced-dimensional spaces at user-defined intervals in the optimization process. We demonstrate the efficacy of this approach on several standard problems from engineering mechanics and scientific machine learning, including two inverse problems, structural topology optimization, and two forward solves with different discretizations. Our results suggest that the learned gradient flows can significantly expedite convergence by capturing critical features of the optimization trajectory while avoiding expensive evaluations of the objective and its gradient.

We are concerned with optimization in a broad sense through the lens of solving variational inequalities (VIs) -- a class of problems that are so general that they cover as particular cases minimization of functions, saddle-point (minimax) problems, Nash equilibrium problems, and many others. The key challenges in our problem formulation are the two-level hierarchical structure and finite-sum representation of the smooth operators in each level. For this setting, we are the first to prove convergence rates and complexity statements for variance-reduced stochastic algorithms approaching the solution of hierarchical VIs in Euclidean and Bregman setups.

Optimizing monotone non-convex functions is a fundamental challenge across machine learning and combinatorial optimization. We introduce and study $γ$-weakly $θ$-up-concavity, a novel first-order condition that characterizes a broad class of such functions. This condition provides a powerful unifying framework, strictly generalizing both DR-submodular functions and One-Sided Smooth (OSS) functions. Our central theoretical contribution demonstrates that $γ$-weakly $θ$-up-concave functions are upper-linearizable: for any feasible point, we can construct a linear surrogate whose gains provably approximate the original non-linear objective. This approximation holds up to a constant factor, namely the approximation coefficient, dependent solely on $γ$, $θ$, and the geometry of the feasible set. This linearizability yields immediate and unified approximation guarantees for a wide range of problems. Specifically, we obtain unified approximation guarantees for offline optimization as well as static and dynamic regret bounds in online settings via standard reductions to linear optimization. Moreover, our framework recovers the optimal approximation coefficient for DR-submodular maximization and significantly improves existing approximation coefficients for OSS optimization, particularly over matroid constraints.

The rapid integration of large language models into newsroom workflows has raised urgent questions about the prevalence of AI-generated content in online media. While computational studies have begun to quantify this phenomenon in English-language outlets, no empirical investigation exists for Turkish news media, where existing research remains limited to qualitative interviews with journalists or fake news detection. This study addresses that gap by fine-tuning a Turkish-specific BERT model (dbmdz/bert-base-turkish-cased) on a labeled dataset of 3,600 articles from three major Turkish outlets with distinct editorial orientations for binary classification of AI-rewritten content. The model achieves 0.9708 F1 score on the held-out test set with symmetric precision and recall across both classes. Subsequent deployment on over 3,500 unseen articles spanning between 2023 and 2026 reveals consistent cross-source and temporally stable classification patterns, with mean prediction confidence exceeding 0.96 and an estimated 2.5 percentage of examined news content rewritten or revised by LLMs on average. To the best of our knowledge, this is the first study to move beyond self-reported journalist perceptions toward empirical, data-driven measurement of AI usage in Turkish news media.

An important goal for personalized diet systems is to improve nutritional quality without compromising convenience or affordability. We present an end-to-end framework that converts dietary standards into complete meals with minimal change. Using the What We Eat in America (WWEIA) intake data for 135,491 meals, we identify 34 interpretable meal archetypes that we then use to condition a generative model and a portion predictor to meet USDA nutritional targets. In comparisons within archetypes, generated meals are better at following recommended daily intake (RDI) targets by 47.0%, while remaining compositionally close to real meals. Our results show that by allowing one to three food substitutions, we were able to create meals that were 10% more nutritious, while reducing costs 19-32%, on average. By turning dietary guidelines into realistic, budget-aware meals and simple swaps, this framework can underpin clinical decision support, public-health programs, and consumer apps that deliver scalable, equitable improvements in everyday nutrition.

Muon-style optimizers leverage Newton-Schulz (NS) iterations to orthogonalize updates, yielding update geometries that often outperform Adam-series methods. However, this orthogonalization discards magnitude information, rendering training sensitive to step-size hyperparameters and vulnerable to high-energy bursts. To mitigate this, we introduce TrasMuon (\textbf{T}rust \textbf{R}egion \textbf{A}daptive \textbf{S}caling \textbf{Muon}). TrasMuon preserves the near-isometric geometry of Muon while stabilizing magnitudes through (i) global RMS calibration and (ii) energy-based trust-region clipping. We demonstrate that while reintroducing adaptive scaling improves optimization efficiency, it typically exacerbates instability due to high-energy outliers. TrasMuon addresses this by defining a trust region based on relative energy ratios, confining updates to a stable zone. Empirical experiments on vision and language models demonstrate that TrasMuon converges faster than baselines. Furthermore, experiments without warmup stages confirm TrasMuon's superior stability and robustness.

1. Many ecological decisions are slowed by the gap between collecting and analysing biodiversity data. Edge computing moves processing closer to the sensor, with edge artificial intelligence (AI) enabling on-device inference, reducing reliance on data transfer and continuous connectivity. In principle, this shifts biodiversity monitoring from passive logging towards autonomous, responsive sensing systems. In practice, however, adoption remains fragmented, with key architectural trade-offs, performance constraints, and implementation challenges rarely reported systematically. 2. Here, we analyse 82 studies published between 2017 and 2025 that implement edge computing for biodiversity monitoring across acoustic, vision, tracking, and multi-modal systems. We synthesise hardware platforms, AI model optimisation, and wireless communication to critically assess how design choices shape ecological inference, deployment longevity, and operational feasibility. 3. Publications increased from 3 in 2017 to 19 in 2025. We identify four system types: (I) TinyML, low-power microcontrollers (MCUs) for single-taxon or rare-event detection; (II) Edge AI, single-board computers (SBCs) for multi-species classification and real-time alerts; (III) Distributed edge AI; and (IV) Cloud AI for retrospective processing pipelines. Each system type represents context-dependent trade-offs among power consumption, computational capability, and communication requirements. 4. Our analysis reveals the evolution of edge computing systems from proof-of-concept to robust, scalable tools. We argue that edge computing offers opportunities for responsive biodiversity management, but realising this potential requires increased collaboration between ecologists, engineers, and data scientists to align model development and system design with ecological questions, field constraints, and ethical considerations.

Pilot studies (PS) are ubiquitous in HCI research. CHI papers routinely reference 'pilot studies', 'pilot tests', or 'preliminary studies' to justify design decisions, verify procedures, or motivate methodological choices. Yet despite their frequency, the role of pilot studies in HCI remains conceptually vague and empirically underexamined. Unlike fields such as medicine, nursing, and education, where pilot and feasibility studies have well-established definitions, guidelines, reporting standards and even a dedicated research journal, the CHI community lacks a shared understanding of what constitutes a pilot study, why they are conducted, and how they should be reported. Many papers reference pilots 'in passing', without details about design, outcomes, or how the pilot informed the main study. This variability suggests a methodological blind spot in our community.

Federated low-rank adaptation (FedLoRA) has facilitated communication-efficient and privacy-preserving fine-tuning of foundation models for downstream tasks. In practical federated learning scenarios, client heterogeneity in system resources and data distributions motivates heterogeneous LoRA ranks across clients. We identify a previously overlooked phenomenon in heterogeneous FedLoRA, termed rank collapse, where the energy of the global update concentrates on the minimum shared rank, resulting in suboptimal performance and high sensitivity to rank configurations. Through theoretical analysis, we reveal the root cause of rank collapse: a mismatch between rank-agnostic aggregation weights and rank-dependent client contributions, which systematically suppresses higher-rank updates at a geometric rate over rounds. Motivated by this insight, we propose raFLoRA, a rank-partitioned aggregation method that decomposes local updates into rank partitions and then aggregates each partition weighted by its effective client contributions. Extensive experiments across classification and reasoning tasks show that raFLoRA prevents rank collapse, improves model performance, and preserves communication efficiency compared to state-of-the-art FedLoRA baselines.

Networks of modern industrial systems are increasingly monitored by distributed sensors, where each system comprises multiple subsystems generating high dimensional time series data. These subsystems are often interdependent, making it important to understand how temporal patterns at one subsystem relate to others. This is challenging in decentralized settings where raw measurements cannot be shared and client observations are heterogeneous. In practical deployments each subsystem (client) operates a fixed proprietary model that cannot be modified or retrained, limiting existing approaches. Nonlinear dynamics further make cross client temporal interdependencies difficult to interpret because they are embedded in nonlinear state transition functions. We present a federated framework for learning temporal interdependencies across clients under these constraints. Each client maps high dimensional local observations to low dimensional latent states using a nonlinear state space model. A central server learns a graph structured neural state transition model over the communicated latent states using a Graph Attention Network. For interpretability we relate the Jacobian of the learned server side transition model to attention coefficients, providing the first interpretable characterization of cross client temporal interdependencies in decentralized nonlinear systems. We establish theoretical convergence guarantees to a centralized oracle and validate the framework through synthetic experiments demonstrating convergence, interpretability, scalability and privacy. Additional real world experiments show performance comparable to decentralized baselines.

Understanding the internal circuits that language models use to solve tasks remains a central challenge in mechanistic interpretability. Most prior work identifies circuits at the task level by averaging across many prompts, implicitly assuming a single stable mechanism per task. We show that this assumption can obscure a crucial source of structure: circuits are prompt-specific, even within a fixed task. Building on attention causal communication (ACC) (Franco & Crovella, 2025), we introduce ACC++, refinements that extract cleaner, lower-dimensional causal signals inside attention heads from a single forward pass. Like ACC, our approach does not require replacement models (e.g., SAEs) or activation patching; ACC++ further improves circuit precision by reducing attribution noise. Applying ACC++ to indirect object identification (IOI) in GPT-2, Pythia, and Gemma 2, we find there is no single circuit for IOI in any model: different prompt templates induce systematically different mechanisms. Despite this variation, prompts cluster into prompt families with similar circuits, and we propose a representative circuit for each family as a practical unit of analysis. Finally, we develop an automated interpretability pipeline that uses ACC++ signals to surface human-interpretable features and assemble mechanistic explanations for prompt families behavior. Together, our results recast circuits as a meaningful object of study by shifting the unit of analysis from tasks to prompts, enabling scalable circuit descriptions in the presence of prompt-specific mechanisms.

Selecting an optimal classification model requires a robust and comprehensive understanding of the performance of the model. This paper provides a tutorial on the PyCM library, demonstrating its utility in conducting deep-dive evaluations of multi-class classifiers. By examining two different case scenarios, we illustrate how the choice of evaluation metrics can fundamentally shift the interpretation of a model's efficacy. Our findings emphasize that a multi-dimensional evaluation framework is essential for uncovering small but important differences in model performance. However, standard metrics may miss these subtle performance trade-offs.

As Large Language Model (LLM) agents become more capable, their coordinated use in the form of multi-agent systems is anticipated to emerge as a practical paradigm. Prior work has examined the safety and misuse risks associated with agents. However, much of this has focused on the single-agent case and/or setups missing basic engineering safeguards such as access control, revealing a scarcity of threat modeling in multi-agent systems. We investigate the security vulnerabilities of a popular multi-agent pattern known as the orchestrator setup, in which a central agent decomposes and delegates tasks to specialized agents. Through red-teaming a concrete setup representative of a likely future use case, we demonstrate a novel attack vector, OMNI-LEAK, that compromises several agents to leak sensitive data through a single indirect prompt injection, even in the \textit{presence of data access control}. We report the susceptibility of frontier models to different categories of attacks, finding that both reasoning and non-reasoning models are vulnerable, even when the attacker lacks insider knowledge of the implementation details. Our work highlights the importance of safety research to generalize from single-agent to multi-agent settings, in order to reduce the serious risks of real-world privacy breaches and financial losses and overall public trust in AI agents.

Robotic foundation models achieve strong generalization by leveraging internet-scale vision-language representations, but their massive computational cost creates a fundamental bottleneck: high inference latency. In dynamic environments, this latency breaks the control loop, rendering powerful models unsafe for real-time deployment. We propose AsyncVLA, an asynchronous control framework that decouples semantic reasoning from reactive execution. Inspired by hierarchical control, AsyncVLA runs a large foundation model on a remote workstation to provide high-level guidance, while a lightweight, onboard Edge Adapter continuously refines actions at high frequency. To bridge the domain gap between these asynchronous streams, we introduce an end-to-end finetuning protocol and a trajectory re-weighting strategy that prioritizes dynamic interactions. We evaluate our approach on real-world vision-based navigation tasks with communication delays up to 6 seconds. AsyncVLA achieves a 40% higher success rate than state-of-the-art baselines, effectively bridging the gap between the semantic intelligence of large models and the reactivity required for edge robotics.

Although foundation models have demonstrated remarkable success in general domains, the application of these models to electroencephalography (EEG) analysis is constrained by substantial data requirements and high parameterization. These factors incur prohibitive computational costs, thereby impeding deployment in resource-constrained clinical environments. Conversely, general-purpose automated machine learning frameworks are often ill-suited for this domain, as exploration within an unbounded programmatic space fails to incorporate essential neurophysiological priors and frequently yields solutions that lack scientific plausibility. To address these limitations, we propose NeuroWeaver, a unified autonomous evolutionary agent designed to generalize across diverse EEG datasets and tasks by reformulating pipeline engineering as a discrete constrained optimization problem. Specifically, we employ a Domain-Informed Subspace Initialization to confine the search to neuroscientifically plausible manifolds, coupled with a Multi-Objective Evolutionary Optimization that dynamically balances performance, novelty, and efficiency via self-reflective refinement. Empirical evaluations across five heterogeneous benchmarks demonstrate that NeuroWeaver synthesizes lightweight solutions that consistently outperform state-of-the-art task-specific methods and achieve performance comparable to large-scale foundation models, despite utilizing significantly fewer parameters.

Multimodal large language models (MLLMs) are changing how Blind and Low Vision (BLV) people access visual information in their daily lives. Unlike traditional visual interpretation tools that provide access through captions and OCR (text recognition through camera input), MLLM-enabled applications support access through conversational assistance, where users can ask questions to obtain goal-relevant details. However, evidence about their performance in the real-world and their implications for BLV people's everyday life remain limited. To address this, we conducted a two-week diary study, where we captured 20 BLV participants' use of an MLLM-enabled visual interpretation application. Although participants rated the visual interpretations of the application as "somewhat trustworthy" (mean=3.76 out of 5, max=very trustworthy) and "somewhat satisfying" (mean=4.13 out of 5, max=very satisfying), the AI often produced incorrect answers (22.2%) or abstained (10.8%) from responding to follow-up requests. Our work demonstrates that MLLMs can improve the accuracy of descriptive visual interpretations, but that supporting everyday use also depends on the "visual assistant" skill -- a set of behaviors for providing goal-directed, reliable assistance. We conclude by proposing the "visual assistant" skill and practical guidelines to help future MLLM-enabled visual interpretation applications better support BLV people's access to visual information.

The ability of machine learning models to store input information in hidden layer vector embeddings, analogous to the concept of `memory', is widely employed but not well characterized. We find that language model embeddings typically contain relatively little input information regardless of data and compute scale during training. In contrast, embeddings from autoencoders trained for input regeneration are capable of nearly perfect memory formation. The substitution of memory embeddings for token sequences leads to substantial computational efficiencies, motivating the introduction of a parallelizable encoder-decoder memory model architecture. Upon causal training these models contain information-poor embeddings incapable of arbitrary information access, but by combining causal and information retention objective functions they learn to form and decode information-rich memories. Training can be further streamlined by freezing a high fidelity encoder followed by a curriculum training approach where decoders first learn to process memories and then learn to additionally predict next tokens. We introduce the perspective that next token prediction training alone is poorly suited for accurate memory formation as the objective itself is non-invertible, motivating the use of combined objective functions for models where the entire input is not exposed.

Large-scale communities of AI agents are becoming increasingly prevalent, creating new environments for agent-agent social interaction. Prior work has examined multi-agent behavior primarily in controlled or small-scale settings, limiting our understanding of emergent social dynamics at scale. The recent emergence of MoltBook, a social networking platform designed explicitly for AI agents, presents a unique opportunity to study whether and how these interactions reproduce core human social mechanisms. We present MoltNet, a large-scale empirical analysis of agent interaction on MoltBook using data collected in early 2026. Grounded in sociological and social-psychological theory, we examine behavior along four dimensions: intent and motivation, norms and templates, incentives and behavioral drift, emotion and contagion. Our analysis revealed that agents strongly respond to social rewards and rapidly converge on community-specific interaction templates, resembling human patterns of incentive sensitivity and normative conformity. However, they are predominantly knowledge-driven rather than persona-aligned, and display limited emotional reciprocity along with weak dialogic engagement, which diverges systematically from human online communities. Together, these results reveal both similarities and differences between artificial and human social systems and provide an empirical foundation for understanding, designing, and governing large-scale agent communities.

The rise of digital technology has dramatically increased the potential for cyberbullying and online abuse, necessitating enhanced measures for detection and prevention, especially among children. This study focuses on detecting abusive obfuscated language in Swahili, a low-resource language that poses unique challenges due to its limited linguistic resources and technological support. Swahili is chosen due to its popularity and being the most widely spoken language in Africa, with over 16 million native speakers and upwards of 100 million speakers in total, spanning regions in East Africa and some parts of the Middle East. We employed machine learning models including Support Vector Machines (SVM), Logistic Regression, and Decision Trees, optimized through rigorous parameter tuning and techniques like Synthetic Minority Over-sampling Technique (SMOTE) to handle data imbalance. Our analysis revealed that, while these models perform well in high-dimensional textual data, our dataset's small size and imbalance limit our findings' generalizability. Precision, recall, and F1 scores were thoroughly analyzed, highlighting the nuanced performance of each model in detecting obfuscated language. This research contributes to the broader discourse on ensuring safer online environments for children, advocating for expanded datasets and advanced machine-learning techniques to improve the effectiveness of cyberbullying detection systems. Future work will focus on enhancing data robustness, exploring transfer learning, and integrating multimodal data to create more comprehensive and culturally sensitive detection mechanisms.

Large language models (LLMs) are increasingly proposed for crisis preparedness and response, particularly for multilingual communication. However, their suitability for high-stakes crisis contexts remains insufficiently evaluated. This work examines the performance of state-of-the-art LLMs and machine translation systems in crisis-domain translation, with a focus on preserving urgency, which is a critical property for effective crisis communication and triaging. Using multilingual crisis data and a newly introduced urgency-annotated dataset covering over 32 languages, we show that both dedicated translation models and LLMs exhibit substantial performance degradation and instability. Crucially, even linguistically adequate translations can distort perceived urgency, and LLM-based urgency classifications vary widely depending on the language of the prompt and input. These findings highlight significant risks in deploying general-purpose language technologies for crisis communication and underscore the need for crisis-aware evaluation frameworks.

An end-to-end autoencoder (AE) framework is developed for downlink non-orthogonal multiple access (NOMA) over Rayleigh fading channels, which learns interference-aware and channel-adaptive super-constellations. While existing works either assume additive white Gaussian noise channels or treat fading channels without a fully end-to-end learning approach, our framework directly embeds the wireless channel into both training and inference. To account for practical channel state information (CSI), we further incorporate limited feedback via both uniform and Lloyd-Max quantization of channel gains and analyze their impact on AE training and bit error rate (BER) performance. Simulation results show that, with perfect CSI, the proposed AE outperforms the existing analytical NOMA schemes. In addition, Lloyd-Max quantization achieves superior BER performance compared to uniform quantization. These results demonstrate that end-to-end AEs trained directly over Rayleigh fading can effectively learn robust, interference-aware signaling strategies, paving the way for NOMA deployment in fading environments with realistic CSI constraints.

Recent vision-language-action (VLA) models can generate plausible end-effector motions, yet they often fail in long-horizon, contact-rich tasks because the underlying hand-object interaction (HOI) structure is not explicitly represented. An embodiment-agnostic interaction representation that captures this structure would make manipulation behaviors easier to validate and transfer across robots. We propose FlowHOI, a two-stage flow-matching framework that generates semantically grounded, temporally coherent HOI sequences, comprising hand poses, object poses, and hand-object contact states, conditioned on an egocentric observation, a language instruction, and a 3D Gaussian splatting (3DGS) scene reconstruction. We decouple geometry-centric grasping from semantics-centric manipulation, conditioning the latter on compact 3D scene tokens and employing a motion-text alignment loss to semantically ground the generated interactions in both the physical scene layout and the language instruction. To address the scarcity of high-fidelity HOI supervision, we introduce a reconstruction pipeline that recovers aligned hand-object trajectories and meshes from large-scale egocentric videos, yielding an HOI prior for robust generation. Across the GRAB and HOT3D benchmarks, FlowHOI achieves the highest action recognition accuracy and a 1.7$\times$ higher physics simulation success rate than the strongest diffusion-based baseline, while delivering a 40$\times$ inference speedup. We further demonstrate real-robot execution on four dexterous manipulation tasks, illustrating the feasibility of retargeting generated HOI representations to real-robot execution pipelines.

Large language models (LLMs) are increasingly deployed in settings where inducing a bias toward a certain topic can have significant consequences, and backdoor attacks can be used to produce such models. Prior work on backdoor attacks has largely focused on a black-box threat model, with an adversary targeting the model builder's LLM. However, in the bias manipulation setting, the model builder themselves could be the adversary, warranting a white-box threat model where the attacker's ability to poison, and manipulate the poisoned data is substantially increased. Furthermore, despite growing research in semantically-triggered backdoors, most studies have limited themselves to syntactically-triggered attacks. Motivated by these limitations, we conduct an analysis consisting of over 1000 evaluations using higher poisoning ratios and greater data augmentation to gain a better understanding of the potential of syntactically- and semantically-triggered backdoor attacks in a white-box setting. In addition, we study whether two representative defense paradigms, model-intrinsic and model-extrinsic backdoor removal, are able to mitigate these attacks. Our analysis reveals numerous new findings. We discover that while both syntactically- and semantically-triggered attacks can effectively induce the target behaviour, and largely preserve utility, semantically-triggered attacks are generally more effective in inducing negative biases, while both backdoor types struggle with causing positive biases. Furthermore, while both defense types are able to mitigate these backdoors, they either result in a substantial drop in utility, or require high computational overhead.

Computation in biological systems is fundamentally energy-constrained, yet standard theories of computation treat energy as freely available. Here, we argue that variational free energy minimization under a Poisson assumption offers a principled path toward an energy-aware theory of computation. Our key observation is that the Kullback-Leibler (KL) divergence term in the Poisson free energy objective becomes proportional to the prior firing rates of model neurons, yielding an emergent metabolic cost term that penalizes high baseline activity. This structure couples an abstract information-theoretic quantity -- the *coding rate* -- to a concrete biophysical variable -- the *firing rate* -- which enables a trade-off between coding fidelity and energy expenditure. Such a coupling arises naturally in the Poisson variational autoencoder (P-VAE) -- a brain-inspired generative model that encodes inputs as discrete spike counts and recovers a spiking form of *sparse coding* as a special case -- but is absent from standard Gaussian VAEs. To demonstrate that this metabolic cost structure is unique to the Poisson formulation, we compare the P-VAE against Grelu-VAE, a Gaussian VAE with ReLU rectification applied to latent samples, which controls for the non-negativity constraint. Across a systematic sweep of the KL term weighting coefficient $β$ and latent dimensionality, we find that increasing $β$ monotonically increases sparsity and reduces average spiking activity in the P-VAE. In contrast, Grelu-VAE representations remain unchanged, confirming that the effect is specific to Poisson statistics rather than a byproduct of non-negative representations. These results establish Poisson variational inference as a promising foundation for a resource-constrained theory of computation.

Large Language Models (LLMs) have shown remarkable potential in scientific domains like retrosynthesis; yet, they often lack the fine-grained control necessary to navigate complex problem spaces without error. A critical challenge is directing an LLM to avoid specific, chemically sensitive sites on a molecule - a task where unconstrained generation can lead to invalid or undesirable synthetic pathways. In this work, we introduce Protect$^*$, a neuro-symbolic framework that grounds the generative capabilities of Large Language Models (LLMs) in rigorous chemical logic. Our approach combines automated rule-based reasoning - using a comprehensive database of 55+ SMARTS patterns and 40+ characterized protecting groups - with the generative intuition of neural models. The system operates via a hybrid architecture: an ``automatic mode'' where symbolic logic deterministically identifies and guards reactive sites, and a ``human-in-the-loop mode'' that integrates expert strategic constraints. Through ``active state tracking,'' we inject hard symbolic constraints into the neural inference process via a dedicated protection state linked to canonical atom maps. We demonstrate this neuro-symbolic approach through case studies on complex natural products, including the discovery of a novel synthetic pathway for Erythromycin B, showing that grounding neural generation in symbolic logic enables reliable, expert-level autonomy.

Does text have an intrinsic curvature? Language is increasingly modeled in curved geometries - hyperbolic spaces for hierarchy, mixed-curvature manifolds for compositional structure - yet a basic scientific question remains unresolved: what does curvature mean for text itself, in a way that is native to language rather than an artifact of the embedding space we choose? We argue that text does indeed have curvature, and show how to detect it, define it, and use it. To this end, we propose Texture, a text-native, word-level discrete curvature signal, and make three contributions. (a) Existence: We provide empirical and theoretical certificates that semantic inference in natural corpora is non-flat, i.e. language has inherent curvature. (b) Definition: We define Texture by reconciling left- and right-context beliefs around a masked word through a Schrodinger bridge, yielding a curvature field that is positive where context focuses meaning and negative where it fans out into competing continuations. (c) Utility: Texture is actionable: it serves as a general-purpose measurement and control primitive enabling geometry without geometric training; we instantiate it on two representative tasks, improving long-context inference through curvature-guided compression and retrieval-augmented generation through curvature-guided routing. Together, our results establish a text-native curvature paradigm, making curvature measurable and practically useful.

The proliferation of data-driven models in weather and climate sciences has marked a significant paradigm shift, with advanced models demonstrating exceptional skill in medium-range forecasting. However, these models are often limited by long-term instabilities, climatological drift, and substantial computational costs during training and inference, restricting their broader application for climate studies. Addressing these limitations, Guan et al. (2024) introduced LUCIE, a lightweight, physically consistent climate emulator utilizing a Spherical Fourier Neural Operator (SFNO) architecture. This model is able to reproduce accurate long-term statistics including climatological mean and seasonal variability. However, LUCIE's native resolution (~300 km) is inadequate for detailed regional impact assessments. To overcome this limitation, we introduce a deep learning-based downscaling framework, leveraging probabilistic diffusion-based generative models with conditional and posterior sampling frameworks. These models downscale coarse LUCIE outputs to 25 km resolution. They are trained on approximately 14,000 ERA5 timesteps spanning 2000-2009 and evaluated on LUCIE predictions from 2010 to 2020. Model performance is assessed through diverse metrics, including latitude-averaged RMSE, power spectrum, probability density functions and First Empirical Orthogonal Function of the zonal wind. We observe that the proposed approach is able to preserve the coarse-grained dynamics from LUCIE while generating fine-scaled climatological statistics at ~28km resolution.

Implicit neural representations (INRs) have emerged as powerful tools for encoding signals, yet dominant MLP-based designs often suffer from slow convergence, overfitting to noise, and poor extrapolation. We introduce FUTON (Fourier Tensor Network), which models signals as generalized Fourier series whose coefficients are parameterized by a low-rank tensor decomposition. FUTON implicitly expresses signals as weighted combinations of orthonormal, separable basis functions, combining complementary inductive biases: Fourier bases capture smoothness and periodicity, while the low-rank parameterization enforces low-dimensional spectral structure. We provide theoretical guarantees through a universal approximation theorem and derive an inference algorithm with complexity linear in the spectral resolution and the input dimension. On image and volume representation, FUTON consistently outperforms state-of-the-art MLP-based INRs while training 2--5$\times$ faster. On inverse problems such as image denoising and super-resolution, FUTON generalizes better and converges faster.

We develop a worst-case complexity theory for stochastically preconditioned stochastic gradient descent (SPSGD) and its accelerated variants under heavy-tailed noise, a setting that encompasses widely used adaptive methods such as Adam, RMSProp, and Shampoo. We assume the stochastic gradient noise has a finite $p$-th moment for some $p \in (1,2]$, and measure convergence after $T$ iterations. While clipping and normalization are parallel tools for stabilizing training of SGD under heavy-tailed noise, there is a fundamental separation in their worst-case properties in stochastically preconditioned settings. We demonstrate that normalization guarantees convergence to a first-order stationary point at rate $\mathcal{O}(T^{-\frac{p-1}{3p-2}})$ when problem parameters are known, and $\mathcal{O}(T^{-\frac{p-1}{2p}})$ when problem parameters are unknown, matching the optimal rates for normalized SGD, respectively. In contrast, we prove that clipping may fail to converge in the worst case due to the statistical dependence between the stochastic preconditioner and the gradient estimates. To enable the analysis, we develop a novel vector-valued Burkholder-type inequality that may be of independent interest. These results provide a theoretical explanation for the empirical preference for normalization over clipping in large-scale model training.

Landauer's principle bounds the heat generated by logical operations, but in practice the thermodynamic cost of computation is dominated by the control systems that implement logic. CMOS gates dissipate energy far above the Landauer bound, while laboratory demonstrations of near-Landauer erasure rely on external measurement or feedback systems whose energy costs exceed that of the logic operation by many orders of magnitude. Here we use simulations to show that a genetic algorithm can program a thermodynamic computer to implement logic operations in which the total heat emitted by the control system is of a similar order of magnitude to that of the information-bearing degrees of freedom. Moreover, the computer can be programmed so that heat is drawn away from the information-bearing degrees of freedom and dissipated within the control unit, suggesting the possibility of computing architectures in which heat management is an integral part of the program design.

Large reasoning models (LRMs) are commonly trained with reinforcement learning (RL) to explore long chain-of-thought reasoning, achieving strong performance at high computational cost. Recent methods add multi-reward objectives to jointly optimize correctness and brevity, but these complex extensions often destabilize training and yield suboptimal trade-offs. We revisit this objective and challenge the necessity of such complexity. Through principled analysis, we identify fundamental misalignments in this paradigm: KL regularization loses its intended role when correctness and length are directly verifiable, and group-wise normalization becomes ambiguous under multiple reward signals. By removing these two items and simplifying the reward to a truncation-based length penalty, we show that the optimization problem reduces to supervised fine-tuning on self-generated data filtered for both correctness and conciseness. We term this simplified training strategy on-policy SFT. Despite its simplicity, on-policy SFT consistently defines the accuracy-efficiency Pareto frontier. It reduces CoT length by up to 80 while maintaining original accuracy, surpassing more complex RL-based methods across five benchmarks. Furthermore, it significantly enhances training efficiency, reducing GPU memory usage by 50% and accelerating convergence by 70%. Our code is available at https://github.com/EIT-NLP/On-Policy-SFT.

Stochastic multi-agent systems are a central modeling framework for autonomous controllers, communication protocols, and cyber-physical infrastructures. In many such systems, however, transition probabilities are only estimated from data and may therefore be partially unknown or subject to perturbations. In this paper, we study the verification of robust strategies in stochastic multi-agent systems with imperfect information, in which coalitions must satisfy a temporal specification while dealing with uncertain system transitions, partial observation, and adversarial agents. By focusing on bounded-memory strategies, we introduce a robust variant of the model-checking problem for a probabilistic, observation-based extension of Alternating-time Temporal Logic. We characterize the complexity of this problem under different notions of perturbation, thereby clarifying the computational cost of robustness in stochastic multi-agent verification and supporting the use of bounded-memory strategies in uncertain environments.

Understanding the causes of software defects is essential for reliable software maintenance and ecosystem stability. However, existing bug datasets do not distinguish between issues originating within a project from those caused by external dependencies or environmental factors. In this paper we present InEx-Bug, a manually annotated dataset of 377 GitHub issues from 103 NPM repositories, categorizing issues as Intrinsic (internal defect), Extrinsic (dependency/environment issue), Not-a-Bug, or Unknown. Beyond labels, the dataset includes rich temporal and behavioral metadata such as maintainer participation, code changes, and reopening patterns. Analyses show Intrinsic bugs resolve faster (median 8.9 vs 10.2 days), are close more often (92% vs 78%), and require code changes more frequently (57% vs 28%) compared to Extrinsic bugs. While Extrinsic bugs exhibit higher reopen rates (12% vs 4%) and delayed recurrence (median 157 vs 87 days). The dataset provides a foundation for further studying Intrinsic and Extrinsic defects in the NPM ecosystem.

Designing cryoprotectant agent (CPA) cocktails for vitrification is challenging because formulations must be concentrated enough to suppress ice formation yet non-toxic enough to preserve cell viability. This tradeoff creates a large, multi-objective design space in which traditional discovery is slow, often relying on expert intuition or exhaustive experimentation. We present a data-efficient framework that accelerates CPA cocktail design by combining high-throughput screening with an active-learning loop based on multi-objective Bayesian optimization. From an initial set of measured cocktails, we train probabilistic surrogate models to predict concentration and viability and quantify uncertainty across candidate formulations. We then iteratively select the next experiments by prioritizing cocktails expected to improve the Pareto front, maximizing expected Pareto improvement under uncertainty, and update the models as new assay results are collected. Wet-lab validation shows that our approach efficiently discovers cocktails that simultaneously achieve high CPA concentrations and high post-exposure viability. Relative to a naive strategy and a strong baseline, our method improves dominated hypervolume by 9.5\% and 4.5\%, respectively, while reducing the number of experiments needed to reach high-quality solutions. In complementary synthetic studies, it recovers a comparably strong set of Pareto-optimal solutions using only 30\% of the evaluations required by the prior state-of-the-art multi-objective approach, which amounts to saving approximately 10 weeks of experimental time. Because the framework assumes only a suitable assay and defined formulation space, it can be adapted to different CPA libraries, objective definitions, and cell lines to accelerate cryopreservation development.

LLM-based agents are becoming increasingly capable, yet their safety lags behind. This creates a gap between what agents can do and should do. This gap widens as agents engage in multi-turn interactions and employ diverse tools, introducing new risks overlooked by existing benchmarks. To systematically scale safety testing into multi-turn, tool-realistic settings, we propose a principled taxonomy that transforms single-turn harmful tasks into multi-turn attack sequences. Using this taxonomy, we construct MT-AgentRisk (Multi-Turn Agent Risk Benchmark), the first benchmark to evaluate multi-turn tool-using agent safety. Our experiments reveal substantial safety degradation: the Attack Success Rate (ASR) increases by 16% on average across open and closed models in multi-turn settings. To close this gap, we propose ToolShield, a training-free, tool-agnostic, self-exploration defense: when encountering a new tool, the agent autonomously generates test cases, executes them to observe downstream effects, and distills safety experiences for deployment. Experiments show that ToolShield effectively reduces ASR by 30% on average in multi-turn interactions. Our code is available at https://github.com/CHATS-lab/ToolShield.

Unmanned aerial vehicles serve as primary sensing platforms for surveillance, traffic monitoring, and disaster response, making aerial object detection a central problem in applied computer vision. Current detectors struggle with UAV-specific challenges: targets spanning only a few pixels, cluttered backgrounds, heavy occlusion, and strict onboard computational budgets. This study introduces LAF-YOLOv10, built on YOLOv10n, integrating four complementary techniques to improve small-object detection in drone imagery. A Partial Convolution C2f (PC-C2f) module restricts spatial convolution to one quarter of backbone channels, reducing redundant computation while preserving discriminative capacity. An Attention-Guided Feature Pyramid Network (AG-FPN) inserts Squeeze-and-Excitation channel gates before multi-scale fusion and replaces nearest-neighbor upsampling with DySample for content-aware interpolation. An auxiliary P2 detection head at 160$\times$160 resolution extends localization to objects below 8$\times$8 pixels, while the P5 head is removed to redistribute parameters. Wise-IoU v3 replaces CIoU for bounding box regression, attenuating gradients from noisy annotations in crowded aerial scenes. The four modules address non-overlapping bottlenecks: PC-C2f compresses backbone computation, AG-FPN refines cross-scale fusion, the P2 head recovers spatial resolution, and Wise-IoU stabilizes regression under label noise. No individual component is novel; the contribution is the joint integration within a single YOLOv10 framework. Across three training runs (seeds 42, 123, 256), LAF-YOLOv10 achieves 35.1$\pm$0.3\% mAP@0.5 on VisDrone-DET2019 with 2.3\,M parameters, exceeding YOLOv10n by 3.3 points. Cross-dataset evaluation on UAVDT yields 35.8$\pm$0.4\% mAP@0.5. Benchmarks on NVIDIA Jetson Orin Nano confirm 24.3 FPS at FP16, demonstrating viability for embedded UAV deployment.

Code review is a critical practice in modern software engineering, helping developers detect defects early, improve code quality, and facilitate knowledge sharing. With the rapid advancement of large language models (LLMs), a growing body of work has explored automated support for code review. However, progress in this area is hindered by the lack of a systematic understanding of existing benchmarks and evaluation practices. Current code review datasets are scattered, vary widely in design, and provide limited insight into what review capabilities are actually being assessed. In this paper, we present a comprehensive survey of code review benchmarks spanning both the Pre-LLM and LLM eras (2015--2025). We analyze 99 research papers (58 Pre-LLM era and 41 LLM era) and extract key metadata, including datasets, evaluation metrics, data sources, and target tasks. Based on this analysis, we propose a multi-level taxonomy that organizes code review research into five domains and 18 fine-grained tasks. Our study reveals a clear shift toward end-to-end generative peer review, increasing multilingual coverage, and a decline in standalone change understanding tasks. We further identify limitations of current benchmarks and outline future directions, including broader task coverage, dynamic runtime evaluation, and taxonomy-guided fine-grained assessment. This survey provides a structured foundation for developing more realistic and comprehensive benchmarks for LLM-based code review.

Vision-Language Models (VLMs) are increasingly used in document processing pipelines to convert flowchart images into structured code (e.g., Mermaid). In production, these systems process arbitrary inputs for which no ground-truth code exists, making output quality difficult to assess. We propose a reference-free evaluation framework that monitors flowchart image-to-code generation quality at inference time, using only the input image and the generated output. The framework introduces two automated metrics: $\text{Recall}{\text{OCR}}$, which estimates content coverage by extracting text from the input image via OCR as a proxy reference, and $\text{Precision}{\text{VE}}$, which detects hallucinated elements through Visual Entailment against the original image. Their harmonic mean, $\text{F1}{\text{OCR-VE}}$, provides a unified quality score. Validation on the FlowVQA dataset shows strong agreement with ground-truth metrics (average Pearson's $r = 0.97$, $0.91$, and $0.94$ for Recall, Precision, and F1, respectively), confirming the framework's reliability as a practical, reference-free alternative for continuous quality monitoring in production settings.

Evaluating moral alignment in agents navigating conflicting, hierarchically structured human norms is a critical challenge at the intersection of AI safety, moral philosophy, and cognitive science. We introduce Morality Chains, a novel formalism for representing moral norms as ordered deontic constraints, and MoralityGym, a benchmark of 98 ethical-dilemma problems presented as trolley-dilemma-style Gymnasium environments. By decoupling task-solving from moral evaluation and introducing a novel Morality Metric, MoralityGym allows the integration of insights from psychology and philosophy into the evaluation of norm-sensitive reasoning. Baseline results with Safe RL methods reveal key limitations, underscoring the need for more principled approaches to ethical decision-making. This work provides a foundation for developing AI systems that behave more reliably, transparently, and ethically in complex real-world contexts.

Multi-agent systems powered by Large Language Models face a critical challenge: agents communicate through natural language, leading to semantic drift, hallucination propagation, and inefficient token consumption. We propose G2CP (Graph-Grounded Communication Protocol), a structured agent communication language where messages are graph operations rather than free text. Agents exchange explicit traversal commands, subgraph fragments, and update operations over a shared knowledge graph, enabling verifiable reasoning traces and eliminating ambiguity. We validate G2CP within an industrial knowledge management system where specialized agents (Diagnostic, Procedural, Synthesis, and Ingestion) coordinate to answer complex queries. Experimental results on 500 industrial scenarios and 21 real-world maintenance cases show that G2CP reduces inter-agent communication tokens by 73%, improves task completion accuracy by 34% over free-text baselines, eliminates cascading hallucinations, and produces fully auditable reasoning chains. G2CP represents a fundamental shift from linguistic to structural communication in multi-agent systems, with implications for any domain requiring precise agent coordination. Code, data, and evaluation scripts are publicly available.

We present Nanbeige4.1-3B, a unified generalist language model that simultaneously achieves strong agentic behavior, code generation, and general reasoning with only 3B parameters. To the best of our knowledge, it is the first open-source small language model (SLM) to achieve such versatility in a single model. To improve reasoning and preference alignment, we combine point-wise and pair-wise reward modeling, ensuring high-quality, human-aligned responses. For code generation, we design complexity-aware rewards in Reinforcement Learning, optimizing both correctness and efficiency. In deep search, we perform complex data synthesis and incorporate turn-level supervision during training. This enables stable long-horizon tool interactions, allowing Nanbeige4.1-3B to reliably execute up to 600 tool-call turns for complex problem-solving. Extensive experimental results show that Nanbeige4.1-3B significantly outperforms prior models of similar scale, such as Nanbeige4-3B-2511 and Qwen3-4B, even achieving superior performance compared to much larger models, such as Qwen3-30B-A3B. Our results demonstrate that small models can achieve both broad competence and strong specialization simultaneously, redefining the potential of 3B parameter models.

Large Language Models are expanding beyond being a tool humans use and into independent agents that can observe an environment, reason about solutions to problems, make changes that impact those environments, and understand how their actions impacted their environment. One of the most common applications of these LLM Agents is in computer programming, where agents can successfully work alongside humans to generate code while controlling programming environments or networking systems. However, with the increasing ability and complexity of these agents comes dangers about the potential for their misuse. A concerning application of LLM agents is in the domain cybersecurity, where they have the potential to greatly expand the threat imposed by attacks such as social engineering. This is due to the fact that LLM Agents can work autonomously and perform many tasks that would normally require time and effort from skilled human programmers. While this threat is concerning, little attention has been given to assessments of the capabilities of LLM coding agents in generating code for social engineering attacks. In this work we compare different LLMs in their ability and willingness to produce potentially dangerous code bases that could be misused by cyberattackers. The result is a dataset of 200 website code bases and logs from 40 different LLM coding agents. Analysis of models shows which metrics of LLMs are more and less correlated with performance in generating spear-phishing sites. Our analysis and the dataset we present will be of interest to researchers and practitioners concerned in defending against the potential misuse of LLMs in spear-phishing.

A key challenge in probabilistic regression is ensuring that predictive distributions accurately reflect true empirical uncertainty. Minimizing overall prediction error often encourages models to prioritize informativeness over calibration, producing narrow but overconfident predictions. However, in safety-critical settings, trustworthy uncertainty estimates are often more valuable than narrow intervals. Realizing the problem, several recent works have focused on post-hoc corrections; however, existing methods either rely on weak notions of calibration (such as PIT uniformity) or impose restrictive parametric assumptions on the nature of the error. To address these limitations, we propose a novel nonparametric re-calibration algorithm based on conditional kernel mean embeddings, capable of correcting calibration error without restrictive modeling assumptions. For efficient inference with real-valued targets, we introduce a novel characteristic kernel over distributions that can be evaluated in $\mathcal{O}(n \log n)$ time for empirical distributions of size $n$. We demonstrate that our method consistently outperforms prior re-calibration approaches across a diverse set of regression benchmarks and model classes.

Machine learning models excel with abundant annotated data, but annotation is often costly and time-intensive. Active learning (AL) aims to improve the performance-to-annotation ratio by using query methods (QMs) to iteratively select the most informative samples. While AL research focuses mainly on QM development, the evaluation of this iterative process lacks appropriate performance metrics. This work reviews eight years of AL evaluation literature and formally introduces the speed-up factor, a quantitative multi-iteration QM performance metric that indicates the fraction of samples needed to match random sampling performance. Using four datasets from diverse domains and seven QMs of various types, we empirically evaluate the speed-up factor and compare it with state-of-the-art AL performance metrics. The results confirm the assumptions underlying the speed-up factor, demonstrate its accuracy in capturing the described fraction, and reveal its superior stability across iterations.

Diffusion Transformers (DiTs) achieve state-of-the-art performance in high-fidelity image and video generation but suffer from expensive inference due to their iterative denoising structure. While prior methods accelerate sampling by caching intermediate features, they rely on static reuse schedules or coarse-grained heuristics, which often lead to temporal drift and cache misalignment that significantly degrade generation quality. We introduce \textbf{AdaCorrection}, an adaptive offset cache correction framework that maintains high generation fidelity while enabling efficient cache reuse across Transformer layers during diffusion inference. At each timestep, AdaCorrection estimates cache validity with lightweight spatio-temporal signals and adaptively blends cached and fresh activations. This correction is computed on-the-fly without additional supervision or retraining. Our approach achieves strong generation quality with minimal computational overhead, maintaining near-original FID while providing moderate acceleration. Experiments on image and video diffusion benchmarks show that AdaCorrection consistently improves generation performance.

Recent advances in mean-field game literature enable the reduction of large-scale multi-agent problems to tractable interactions between a representative agent and a population distribution. However, existing approaches typically assume a fixed initial population distribution and fully rational agents, limiting robustness under distributional uncertainty and cognitive constraints. We address these limitations by introducing risk aversion with respect to the initial population distribution and by incorporating bounded rationality to model deviations from fully rational decision-making agents. The combination of these two elements yields a new and more general equilibrium concept, which we term the mean-field risk-averse quantal response equilibrium (MF-RQE). We establish existence results and prove convergence of fixed-point iteration and fictitious play to MF-RQE. Building on these insights, we develop a scalable reinforcement learning algorithm for scenarios with large state-action spaces. Numerical experiments demonstrate that MF-RQE policies achieve improved robustness relative to classical mean-field approaches that optimize expected cumulative rewards under a fixed initial distribution and are restricted to entropy-based regularizers.

Automated image captioning using the content from the image is very appealing when done by harnessing the capability of computer vision and natural language processing. Extensive research has been done in the field with a major focus on the English language which gives the scope for further developments in the same with consideration of popular foreign languages. This research utilizes distinct models for translating the image caption into Hindi, the fourth most popular language across the world. Exploring the multi-modal architectures this research comprises local visual features, global visual features, attention mechanisms, and pre-trained models. Using google cloud translator on the image dataset from Flickr8k, Hindi image descriptions have been generated. Pre-trained CNNs like VGG16, ResNet50, and Inception V3 helped in retrieving image characteristics, while the uni-directional and bi-directional techniques of text encoding are used for the text encoding process. An additional Attention layer helps to generate a weight vector and, by multiplying it, combine image characteristics from each time step into a sentence-level feature vector. Bilingual evaluation understudy scores are used to compare the research outcome. Many experiments that serve as a baseline are done for the comparative analysis of the research. An image with a score of BLEU-1 is considered sufficient, whereas one with a score of BLEU-4 is considered to have fluid image captioning. For both BLEU scores, the attention-based bidirectional LSTM with VGG16 produced the best results of 0.59 and 0.19 respectively. The experiments conclude that researchs ability to produce relevant, semantically accurate image captions in Hindi. The research accomplishes the goals and future research can be guided by this research model.

Finite automata (FA) are a fundamental computational abstraction that is widely used in practice for various tasks in computer science, linguistics, biology, electrical engineering, and artificial intelligence. Given an input word, an FA maps the word to a result, in the simple case "accept" or "reject", but in general to one of a finite set of results. A question that then arises is: why? Another question is: how can we modify the input word so that it is no longer accepted? One may think that the automaton itself is an adequate explanation of its behaviour, but automata can be very complex and difficult to make sense of directly. In this work, we investigate how to explain the behaviour of an FA on an input word in terms of the word's characters. In particular, we are interested in minimal explanations: what is the minimal set of input characters that explains the result, and what are the minimal changes needed to alter the result? In this paper, we propose an efficient method to determine all minimal explanations for the behaviour of an FA on a particular word. This allows us to give unbiased explanations about which input features are responsible for the result. Experiments show that our approach scales well, even when the underlying problem is challenging.

Brick kilns are a major source of air pollution and forced labor in South Asia, yet large-scale monitoring remains limited by sparse and outdated ground data. We study brick kiln detection at scale using high-resolution satellite imagery and curate a multi city zoom-20 (0.149 meters per pixel) resolution dataset comprising over 1.3 million image tiles across five regions in South and Central Asia. We propose ClimateGraph, a region-adaptive graph-based model that captures spatial and directional structure in kiln layouts, and evaluate it against established graph learning baselines. In parallel, we assess a remote sensing based detection pipeline and benchmark it against recent foundation models for satellite imagery. Our results highlight complementary strengths across graph, foundation, and remote sensing approaches, providing practical guidance for scalable brick kiln monitoring from satellite imagery.

Text-to-image models have made significant strides, producing impressive results in generating images from textual descriptions. However, creating a scalable pipeline for deploying these models in production remains a challenge. Achieving the right balance between automation and human feedback is critical to maintain both scale and quality. While automation can handle large volumes, human oversight is still an essential component to ensure that the generated images meet the desired standards and are aligned with the creative vision. This paper presents a new pipeline that offers a fully automated, scalable solution for generating marketing images of commercial products using text-to-image models. The proposed system maintains the quality and fidelity of images, while also introducing sufficient creative variation to adhere to marketing guidelines. By streamlining this process, we ensure a seamless blend of efficiency and human oversight, achieving a $30.77\%$ increase in marketing object fidelity using DINOV2 and a $52.00\%$ increase in human preference over the generated outcome.

Small datasets like MNIST have historically been instrumental in advancing machine learning research by providing a controlled environment for rapid experimentation and model evaluation. However, their simplicity often limits their utility for distinguishing between advanced neural network architectures. To address these challenges, Greydanus et al. introduced the MNIST-1D dataset, a one-dimensional adaptation of MNIST designed to explore inductive biases in sequential data. This dataset maintains the advantages of small-scale datasets while introducing variability and complexity that make it ideal for studying advanced architectures. In this paper, we extend the exploration of MNIST-1D by evaluating the performance of Residual Networks (ResNet), Temporal Convolutional Networks (TCN), and Dilated Convolutional Neural Networks (DCNN). These models, known for their ability to capture sequential patterns and hierarchical features, were implemented and benchmarked alongside previously tested architectures such as logistic regression, MLPs, CNNs, and GRUs. Our experimental results demonstrate that advanced architectures like TCN and DCNN consistently outperform simpler models, achieving near-human performance on MNIST-1D. ResNet also shows significant improvements, highlighting the importance of leveraging inductive biases and hierarchical feature extraction in small structured datasets. Through this study, we validate the utility of MNIST-1D as a robust benchmark for evaluating machine learning architectures under computational constraints. Our findings emphasize the role of architectural innovations in improving model performance and offer insights into optimizing deep learning models for resource-limited environments.

Automated warehouses execute millions of stow operations, where robots place objects into storage bins. For these systems it is valuable to anticipate how a bin will look from the current observations and the planned stow behavior before real execution. We propose FOREST, a stow-intent-conditioned world model that represents bin states as item-aligned instance masks and uses a latent diffusion transformer to predict the post-stow configuration from the observed context. Our evaluation shows that FOREST substantially improves the geometric agreement between predicted and true post-stow layouts compared with heuristic baselines. We further evaluate the predicted post-stow layouts in two downstream tasks, in which replacing the real post-stow masks with FOREST predictions causes only modest performance loss in load-quality assessment and multi-stow reasoning, indicating that our model can provide useful foresight signals for warehouse planning.

Single-cell RNA-seq (scRNA-seq) enables atlas-scale profiling of complex tissues, revealing rare lineages and transient states. Yet, assigning biologically valid cell identities remains a bottleneck because markers are tissue- and state-dependent, and novel states lack references. We present CellMaster, an AI agent that mimics expert practice for zero-shot cell-type annotation. Unlike existing automated tools, CellMaster leverages LLM-encoded knowledge (e.g., GPT-4o) to perform on-the-fly annotation with interpretable rationales, without pre-training or fixed marker databases. Across 9 datasets spanning 8 tissues, CellMaster improved accuracy by 7.1% over best-performing baselines (including CellTypist and scTab) in automatic mode. With human-in-the-loop refinement, this advantage increased to 18.6%, with a 22.1% gain on subtype populations. The system demonstrates particular strength in rare and novel cell states where baselines often fail. Source code and the web application are available at \href{https://github.com/AnonymousGym/CellMaster}{https://github.com/AnonymousGym/CellMaster}.

Decades of engineering drawings and technical records remain locked in legacy archives with inconsistent or missing metadata, making retrieval difficult and often manual. We present Blueprint, a layout-aware multimodal retrieval system designed for large-scale engineering repositories. Blueprint detects canonical drawing regions, applies region-restricted VLM-based OCR, normalizes identifiers (e.g., DWG, part, facility), and fuses lexical and dense retrieval with a lightweight region-level reranker. Deployed on ~770k unlabeled files, it automatically produces structured metadata suitable for cross-facility search. We evaluate Blueprint on a 5k-file benchmark with 350 expert-curated queries using pooled, graded (0/1/2) relevance judgments. Blueprint delivers a 10.1% absolute gain in Success@3 and an 18.9% relative improvement in nDCG@3 over the strongest vision-language baseline}, consistently outperforming across vision, text, and multimodal intents. Oracle ablations reveal substantial headroom under perfect region detection and OCR. We release all queries, runs, annotations, and code to facilitate reproducible evaluation on legacy engineering archives.

Unstructured documents like PDFs contain valuable structured information, but downstream systems require this data in reliable, standardized formats. LLMs are increasingly deployed to automate this extraction, making accuracy and reliability paramount. However, progress is bottlenecked by two gaps. First, no end-to-end benchmark evaluates PDF-to-JSON extraction under enterprise-scale schema breadth. Second, no principled methodology captures the semantics of nested extraction, where fields demand different notions of correctness (exact match for identifiers, tolerance for quantities, semantic equivalence for names), arrays require alignment, and omission must be distinguished from hallucination. We address both gaps with ExtractBench, an open-source benchmark and evaluation framework for PDF-to-JSON structured extraction. The benchmark pairs 35 PDF documents with JSON Schemas and human-annotated gold labels across economically valuable domains, yielding 12,867 evaluatable fields spanning schema complexities from tens to hundreds of fields. The evaluation framework treats the schema as an executable specification: each field declares its scoring metric. Baseline evaluations reveal that frontier models (GPT-5/5.2, Gemini-3 Flash/Pro, Claude 4.5 Opus/Sonnet) remain unreliable on realistic schemas. Performance degrades sharply with schema breadth, culminating in 0% valid output on a 369-field financial reporting schema across all tested models. We release ExtractBench at https://github.com/ContextualAI/extract-bench.

Do Large Language Models (LLMs) possess a Theory of Mind (ToM)? Research into this question has focused on evaluating LLMs against benchmarks and found success across a range of social tasks. However, these evaluations do not test for the actual representations posited by ToM: namely, a causal model of mental states and behavior. Here, we use a cognitively-grounded definition of ToM to develop and test a new evaluation framework. Specifically, our approach probes whether LLMs have a coherent, domain-general, and consistent model of how mental states cause behavior -- regardless of whether that model matches a human-like ToM. We find that even though LLMs succeed in approximating human judgments in a simple ToM paradigm, they fail at a logically equivalent task and exhibit low consistency between their action predictions and corresponding mental state inferences. As such, these findings suggest that the social proficiency exhibited by LLMs is not the result of a domain-general or consistent ToM.

Macroscopic traffic safety modeling aims to identify critical risk factors for regional crashes, thereby informing targeted policy interventions for safety improvement. However, current approaches rely heavily on static sociodemographic and infrastructure metrics, frequently overlooking the impacts from drivers' visual perception of driving environment. Although visual environment features have been found to impact driving and traffic crashes, existing evidence remains largely observational, failing to establish the robust causality for traffic policy evaluation under complex spatial environment. To fill these gaps, we applied semantic segmentation on Google Street View imageries to extract visual environmental features and proposed a Double Machine Learning framework to quantify their causal effects on regional crashes. Meanwhile, we utilized SHAP values to characterize the nonlinear influence mechanisms of confounding variables in the models and applied causal forests to estimate conditional average treatment effects. Leveraging crash records from the Miami metropolitan area, Florida, and 220,000 street view images, evidence shows that greenery proportion exerts a significant and robust negative causal effect on traffic crashes (Average Treatment Effect = -6.38, p = 0.005). This protective effect exhibits spatial heterogeneity, being most pronounced in densely populated and socially vulnerable urban cores. While greenery significantly mitigates angle and rear-end crashes, its protective benefit for vulnerable road users (VRUs) remains limited. Our findings provide causal evidence for greening as a potential safety intervention, prioritizing hazardous visual environments while highlighting the need for distinct design optimizations to protect VRUs.

We introduce potential-energy gating, a method for robust state estimation in systems governed by double-well stochastic dynamics. The observation noise covariance of a Bayesian filter is modulated by the local value of a known or assumed potential energy function: observations are trusted when the state is near a potential minimum and progressively discounted as it approaches the barrier separating metastable wells. This physics-based mechanism differs from statistical robust filters, which treat all state-space regions identically, and from constrained filters, which bound states rather than modulating observation trust. The approach is especially relevant in non-ergodic or data-scarce settings where only a single realization is available and statistical methods alone cannot learn the noise structure. We implement gating within Extended, Unscented, Ensemble, and Adaptive Kalman filters and particle filters, requiring only two additional hyperparameters. Monte Carlo benchmarks (100 replications) on a Ginzburg-Landau double-well with 10% outlier contamination show 57-80% RMSE improvement over the standard Extended Kalman Filter, all statistically significant (p < 10^{-15}, Wilcoxon test). A naive topological baseline using only well positions achieves 57%, confirming that the continuous energy landscape adds ~21 percentage points. The method is robust to misspecification: even with 50% parameter errors, improvement never falls below 47%. Comparing externally forced and spontaneous Kramers-type transitions, gating retains 68% improvement under noise-induced transitions whereas the naive baseline degrades to 30%. As an empirical illustration, we apply the framework to Dansgaard-Oeschger events in the NGRIP delta-18O ice-core record, estimating asymmetry gamma = -0.109 (bootstrap 95% CI: [-0.220, -0.011]) and showing that outlier fraction explains 91% of the variance in filter improvement.

Visual navigation models often struggle in real-world dynamic environments due to limited robustness to the sim-to-real gap and the difficulty of training policies tailored to target deployment environments (e.g., households, restaurants, and factories). Although real-to-sim navigation simulation using 3D Gaussian Splatting (GS) can mitigate these challenges, prior GS-based works have considered only static scenes or non-photorealistic human obstacles built from simulator assets, despite the importance of safe navigation in dynamic environments. To address these issues, we propose ReaDy-Go, a novel real-to-sim simulation pipeline that synthesizes photorealistic dynamic scenarios in target environments by augmenting a reconstructed static GS scene with dynamic human GS obstacles, and trains navigation policies using the generated datasets. The pipeline provides three key contributions: (1) a dynamic GS simulator that integrates static scene GS with a human animation module, enabling the insertion of animatable human GS avatars and the synthesis of plausible human motions from 2D trajectories, (2) a navigation dataset generation framework that leverages the simulator along with a robot expert planner designed for dynamic GS representations and a human planner, and (3) robust navigation policies to both the sim-to-real gap and moving obstacles. The proposed simulator generates thousands of photorealistic navigation scenarios with animatable human GS avatars from arbitrary viewpoints. ReaDy-Go outperforms baselines across target environments in both simulation and real-world experiments, demonstrating improved navigation performance even after sim-to-real transfer and in the presence of moving obstacles. Moreover, zero-shot sim-to-real deployment in an unseen environment indicates its generalization potential. Project page: https://syeon-yoo.github.io/ready-go-site/.

4D mmWave radar provides weather-robust, velocity-aware measurements and is more cost-effective than LiDAR. However, radar-only 3D detection still trails LiDAR-based systems because radar point clouds are sparse, irregular, and often corrupted by multipath noise, yielding weak and unstable geometry. We present HyperDet, a detector-agnostic radar-only 3D detection framework that constructs a task-aware hyper 4D radar point cloud for standard LiDAR-oriented detectors. HyperDet aggregates returns from multiple surround-view 4D radars over consecutive frames to improve coverage and density, then applies geometry-aware cross-sensor consensus validation with a lightweight self-consistency check outside overlap regions to suppress inconsistent returns. It further integrates a foreground-focused diffusion module with training-time mixed radar-LiDAR supervision to densify object structures while lifting radar attributes (e.g., Doppler, RCS); the model is distilled into a consistency model for single-step inference. On MAN TruckScenes, HyperDet consistently improves over raw radar inputs with VoxelNeXt and CenterPoint, partially narrowing the radar-LiDAR gap. These results show that input-level refinement enables radar to better leverage LiDAR-oriented detectors without architectural modifications.

The understanding of fractional Chern insulators (FCIs) has been deeply guided by band topology and quantum geometry. Here, we introduce a real-space theoretical framework in which FCIs are understood in terms of composite bosons, local objects consisting of electrons bound to their energetically excluded surrounding orbitals. The central element of our framework is the construction of a radially ordered set of maximally localized basis for Chern bands without requiring continuous rotational symmetry. Within this basis, the complex many-body problem simplifies to a real-space organizing principle: a stable FCI occurs if the orbitals excluded around central electrons are those maximizing the two-body interaction energy. We validate this with direct numerical evidence for composite boson formation in the Haldane model, demonstrating that our criterion reliably characterizes FCIs. Importantly, our analysis illustrates that the composite boson framework bridges the fractional quantum Hall effect in continuum and lattice paradigms, providing a unified and intuitive real-space interpretation for distinct correlated phases. It thus establishes a foundation for diagnosing and guiding the design of both Abelian and non-Abelian topologically ordered phases across distinct platforms.

We study caging/trapping in Sokoban-type models, featuring a random walker moving through a disordered medium of obstacles and capable of pushing some obstacles blocking its path. In one-dimension, we allow the walker to push up to an arbitrary $N_{\rm P}$ number of obstacles. For $N_{\rm P}\gg 1$, we use large-deviation theory to show that the survival probability to remain uncaged exhibits crossover from an exponential decay with time at intermediate times to a stretched-exponential decay at long times, with an exponent $1/3$ independent of $N_{\rm P}$. The long-time exponent matches the Balagurov--Vaks--Donsker--Varadhan (BVDV) theory of the classical trapping problem, while the exponential decay is qualitatively distinct from the Rosenstock's intermediate-time theory for classical trapping. Similarly, in two dimensions, numerical simulations reveal that both the Sokoban model and its generalized version exhibit long-time stretched-exponential relaxation with exponent $1/2$, again consistent with the BVDV theory. Finally, in two dimensions, we find that the mean trap size is nonmonotonic in $ρ$: it is small at both low and high densities, but reaches a peak at a characteristic density $ρ_*$. We estimate $ρ_* \approx 0.55$ for the Sokoban model and $ρ_* \approx 0.675$ for the generalized Sokoban model.

Controlling nonlinear effects in micro- and nano-electro-mechanical systems is essential for unlocking their full potential in sensing, signal processing, and frequency control. In this study, we develop a voltage-dependent Hamiltonian framework for a nanoelectromechanical resonator with two strongly coupled vibrational modes, representative of a nanostring platform. The mode frequencies and couplings of the system are tuned electrostatically using a DC voltage, which also controls the strength of the interactions. Our theoretical model reproduces the experimentally observed avoided crossing in the absence of an AC drive and generates tunable frequency-comb spectra when a parametric drive is applied. By scanning the DC voltage, we generate a phase diagram that links comb formation and sharp regime boundaries to underlying bifurcations, multi-stability, and attractor switching. Phase-resolved diagnostics based on a Kuramoto order parameter, together with autocorrelation and Poincaré analyses, quantify coherence and critical slowing down near these transitions. We further explore the relationship between nonlinear coupling, parametric excitation, and stability transitions within a single device of experimental relevance and establish a dynamical framework for engineering nanoelectromechanical resonators that offer enhanced tunability, functionality, and a predictive link to experimental outcomes.

For a class of translation-invariant free-fermion systems (including those with uniform nearest neighbor hopping) on a $d$-dimensional $L \times \cdots \times L$ hypercubic lattice, we prove that, starting from an arbitrary pure initial state, the system equilibrates with respect to the coarse-grained density within a timescale of order $L$. This scaling is optimal, since there exist initial states whose equilibration requires time of order $L$. Our result establishes $O(L)$ as the equilibration timescale, as is expected in normal macroscopic systems with a conserved quantity, such as total number of particles.

We report a 139La-NMR study of Ba2La2CoTe2O12, S = 1/2 equilateral triangular-lattice antiferromagnet with easy-plane anisotropy at low temperatures. This compound undergoes a magnetic phase transition at TN = 3.26 K into an ordered state with the 120 degree spin structure. Under magnetic fields above 3T, TN splits into TN1 and TN2, which correspond to the transitions from the paramagnetic phase to the up-up-down (uud) phase and from the uud phase to the triangular coplanar phase, respectively. The NMR spin-lattice relaxation rate 1/T1 exhibits a critical divergence at TN1, indicating the onset of long-range magnetic order. At TN2, the NMR-linewidth measured at 5.4 T exhibits an anomalous decrease, which we attribute to a change in the spin structure from the uud to the triangular coplanar phase.

We propose a novel type of exceptional points, dubbed interaction-enabled $n$-fold exceptional points [EP$n$s ($n=2,3$)] -- EP$n$s protected by topology that are prohibited at the non-interacting level. Specifically, we demonstrate that both bosonic and fermionic systems host such interaction-enabled EP$n$s ($n=2,3$) in parameter space that are protected by charge U(1), pseudo-spin-parity, and $PT$ symmetries. The interaction-enabled EP2s are protected by zero-dimensional topology and give rise to qualitative changes in the loss rate, an experimentally measurable quantity for cold atoms. Furthermore, we reveal that interactions enable EP3s protected by one-dimensional topology beyond the point-gap topological classifications, suggesting the potential presence of a broader class of interaction-enabled non-Hermitian degeneracies.

The interplay of topology, disorder, and non-Hermiticity gives rise to phenomena beyond the conventional classification of quantum phases. We propose a one-dimensional non-Hermitian Su-Schrieffer-Heeger model with quasiperiodically modulated nonreciprocal intracell hopping. We show that quasiperiodic modulation can substantially enhance topological robustness and, remarkably, induce a non-Hermitian topological Anderson insulator (TAI) phase. Beyond the topological transition, increasing nonreciprocity drives a cascade of localization transitions in which all bulk eigenstates evolve from extended to multifractal critical and ultimately to localized states. Strikingly, the extended-to-critical transition coincides exactly with a real-complex spectral transition. We establish complete phase diagrams and derive exact analytical boundaries for both topological and localization transitions, uncovering an unanticipated coexistence of TAI and multifractal critical phases. Finally, we propose a feasible implementation in topolectrical circuits. Our results reveal a new paradigm for the cooperative effects of topology, quasiperiodicity, and non-Hermiticity.

The nitrogen-vacancy (NV) center in diamond exists in different charge states with distinct photoluminescence properties, which are sensitive to the nanoscale electrochemical environment. Hence, the NV charge state is emerging as a powerful all-optical platform for nanoscale sensing and imaging. Although significant progress has been made in engineering near-surface NV centers in bulk diamond, controlling the NV charge state in fluorescent nanodiamonds (FNDs) has proven challenging, limiting the sensitivity and reliability of FND-based charge state sensing. Here, we demonstrate reliable, reversible switching between the fluorescent NV$^0$ and non-fluorescent NV$^+$ charge states in sub-30 nm FNDs via surface oxidation and hydrogenation, respectively, for single particles and particle powder. In aqueous electrochemical cells, we demonstrate voltage and ion concentration imaging based on the NV charge state in self-assembled FND layers on transparent substrates. Applied voltages reliably modulate the FND PL with a sensitivity of up to 16 mV Hz$^{-1/2}$. Importantly, FND PL is also modulated by local changes in salt concentration with a sensitivity of up to 1.8% per millimolar NaCl, enabling all-optical imaging of ion concentration gradients at the microscale. Our results represent a significant step toward realizing fast, stable, and scalable nanoscale charge- and voltage-imaging technologies with sub-micrometer spatial resolution.

The nucleation of quasicrystals remains a fundamental puzzle, primarily due to the absence of a periodic translational template. Here, we demonstrate that phasons - hidden degrees of freedom unique to quasiperiodic order - drive diverse nucleation pathways in icosahedral quasicrystals (IQCs). Combining a Landau free-energy model with the spring pair method, we compute distinct critical nuclei and their corresponding minimum energy paths. At low temperatures, a direct, symmetry-preserving pathway dominates. In contrast, higher temperatures promote a "symmetry detour" that reduces the nucleation barrier via a lower-symmetry critical nucleus. Remarkably, while the resulting bulk IQCs exhibit distinct real-space symmetries, they remain thermodynamically degenerate with identical diffraction patterns. We resolve this paradox within the high-dimensional projection framework, showing that phason shifts modulate real-space symmetry without altering bulk thermodynamics. Our findings establish phasons as the structural origin of pathway diversity, offering a new physical picture for the emergence of quasiperiodic order.

Can geometry alone reshape equilibrium? Cascaded nanofluidic chambers show complex accumulation patterns, traditionally attributed to geometric diode effects. We use 3D molecular dynamics to decouple funnel rectification from boundary reflection. Simulations with argon parameters (r = 0.19 nm) reveal a striking "reverse" rectification in a 2-chamber setup: the narrow side accumulates over 5x more particles (N_1/N_0 = 5.37 +/- 0.01, p < 0.0001). In a 10-chamber argon cascade, this effect drives massive downstream accumulation. A symmetric control (w_L = w_R) eliminates the gradient, confirming that funnel asymmetry - not boundary/edge effects - is the primary driver in the ballistic regime. By contrast, the super-atom regime is dominated by boundary reflection. Our results challenge standard entropic transport theory and provide design rules for passive, geometry-driven density gradients - no pump, no drive.

In contrast to eigenvalue-based approaches, we formulate the bulk-boundary correspondence for two-dimensional non-Hermitian quadratic lattice Hamiltonians in terms of Toeplitz operators and singular values, which correctly capture the stability, localization, and scaling of edge and corner modes. We show that singular values, rather than eigenvalues, provide the only stable foundation for topological protection in non-Hermitian systems because they remain robust under translational-symmetry-breaking perturbations that destabilize the eigenvalue spectrum, rendering it unsuitable for topological classification. Building on Toeplitz operator theory, we establish general results for non-Hermitian Hamiltonians defined on half and quarter planes, relating the topological indices of the associated Toeplitz operators to the number of finite-size singular values that are separated from the bulk singular-value spectrum and vanish in the thermodynamic limit. This yields a precise bulk-boundary correspondence for edge and corner modes, including higher-order topological phases, without requiring crystalline symmetries. We illustrate our general results with detailed examples exhibiting topologically protected families of edge states, coexisting edge and corner modes, and phases with both gapped bulk and edges supporting only stable corner modes. The latter is exemplified by a non-Hermitian generalization of the Benalcazar-Bernevig-Hughes model.

Predicting the rheological behavior of associative polymers bridging colloidal particles into transient networks is fundamentally challenging because the coupled spatiotemporal scales prevent efficient molecular-fidelity modeling. We address this through a novel, unified multiscale simulation framework for telechelic polymer-colloid suspensions integrating: explicit-chain Brownian dynamics resolving polymer-particle association kinetics; active learning metamodels compressing kinetics into efficient surrogates; and Population Balance-Brownian Dynamics (Pop-BD) computing network-scale dynamics from metamodel predictions. Validated against explicit-chain Brownian dynamics, our framework accurately reproduces time-and frequency-dependent stress relaxation moduli, enabling simulations of larger systems over longer timescales. Systematic investigations reveal that network connectivity exhibits critical transitions at specific chain-to-particle ratios, with bond density and lifetime correlating to enhanced relaxation times and moduli. Higher particle volume fractions yield more persistent bonds and slower relaxation. This framework connects chain-level dynamics to macroscopic rheology, enabling computationally efficient rational design of associative colloidal materials for waterborne coatings and soft-matter applications.

We investigate the influence of dynamic heterogeneity on the spectral shape of structural relaxation in van der Waals liquids above the melting point by means of depolarized dynamic light scattering. To this end, we study optically anisotropic probe molecules both in the bulk and when diluted in an optically isotropic solvent. Strikingly, the relaxation shape of the probe molecules in dilution is indistinguishable from that of the pure liquid composed of the probe molecules. By contrast, when explicit dynamic heterogeneity is introduced, e.g., through internal degrees of freedom or a distribution of probe molecule sizes, the relaxation shape becomes sensitive to the solvent concentration. These findings indicate that dynamic heterogeneity has a negligible influence on the rotational dynamics of single component van der Waals liquids above the melting point, despite the pronounced non-exponential character of their relaxation shape.

This paper considers the propagation of excitons in linear chains of QDs with a side defect, located on a dielectric substrate. This configuration is suitable for spatially selective excitation of the system by pulsed optical radiation through the side defect. The dynamics of excitation in the chain is governed by structural disorder, caused by technological variations in the parameters of the QDs themselves and their mutual arrangement. To describe the quantum properties of excitons in the QD chain, a model Hamiltonian is used, taking into account the coupling of neighboring QDs due to dipole-dipole interaction. The localization of stationary states is calculated depending on the magnitude of disorder and the chain length. A criterion is introduced that determines the boundaries of the phase transition from the localized to the delocalized excitation phase. The dynamics of exciton transfer along the QD chain is calculated depending on its length and degree of disorder for linear excitation of the system by a laser pulse. It is shown that dynamic localization emerging in the system corresponds to the stationary localization properties of the states of the chain with a side defect.

Overheating is a critical bottleneck limiting the performance and reliability of next-generation high-power and high-frequency electronics. Interfacial thermal resistance constitutes a significant portion of the total thermal resistance. In this study, we report an ultrahigh thermal boundary conductance (TBC) of approximately 800 MW/m2-K at the atomically-sharp AlN-SiC interface, achieved through an ion implantation-enhanced nucleation epitaxy technique. This value is among the highest TBC values reported for semiconductor interfaces, confirmed by structural characterizations which show an ultrahigh-quality interface. Atomistic Green Function calculations reveal that elastic phonon transmission dominates the interface, with nearly half of the acoustic modes (0-15 THz) exhibiting near-unity transmission due to the atomically sharp structure. Furthermore, using high-energy-resolution electron energy loss spectroscopy, we probe vibrational properties with nanometer spatial resolution and identify unique interfacial phonon modes connecting the mismatched phonon spectra, confirmed by molecular dynamics simulations. The ultrahigh TBC is attributed to both the high elastic phonon transmission due to the high quality interfaces and the inelastic phonon scattering channel due to interfacial phonon modes. These findings not only advance the fundamental understanding of interfacial thermal transport but also provide a pathway for effective thermal management in emerging electronic devices.

This work explores the lift force experienced by a particle moving in a viscous fluid near a liquid-liquid interface. The lift force is induced by the interaction between the viscous flow generated by the particle's motion and the deformation of the soft interface. The factors influencing the lift force including the velocity, the viscosity, and the sphere radius, and the separation distance were systematically studied. The experiments demonstrate that the lift force intensifies as the particle approaches the interface, and saturates at shorter distances. These findings are consistent with predictions made using soft lubrication theory and numerical calculations, providing strong validation for the theoretical framework.

In this work, we compute the defect relative entropy between topological defects in the symmetric product orbifold CFT $\mathrm{Sym}^N(M) = M^{\otimes N}/S_N$. Our analysis covers two distinct classes of defects: universal defects, which realize the $\mathrm{Rep}(S_N)$ non-invertible symmetry, and non-universal defects. We show that the defect relative entropy reduces to a Kullback--Leibler (KL) divergence. The resulting expression decomposes naturally into two contributions: one governed by characters of the symmetric group $S_N$, and the other controlled by modular $S$-matrix elements of the seed RCFT. Remarkably, both sets of data appear as probability distributions, yielding an information-theoretic interpretation of permutation group data and modular data within the symmetric orbifold. The structure of the divergence depends sensitively on the defect class. For universal defects, only the permutation group data contributes; for maximally fractional defects, both permutation and modular data enter and together define the relevant probability distributions. This feature suggests that the maximally fractional defect can be understood as a kind of product of the RCFT defect and the symmetric orbifold defect.

The efficient manipulation of thermodynamic states within the finite time is fundamentally constrained by the intrinsic dissipative cost. While the slow-driving regime is well-characterized by a universal $1/τ$-scaling of irreversibility, the physics governing fast, non-adiabatic transitions remains elusive. Here, we propose the polytropic steering protocols that provide an exact analytical bridge between the isothermal and adiabatic limits for Brownian particles far-from-equilibrium. We demonstrate that for any protocol duration $τ$, the system can be precisely steered along a prescribed polytropic trajectory, revealing a striking non-monotonic dependence of irreversibility on the driving rate. Contrary to the near-equilibrium paradigm where faster driving necessitates higher energetic costs, we identify a most-irreversible timescale, beyond which dissipation is anomalously suppressed by rapid driving. By mapping these protocols onto a broad class of controllable thermodynamic cycle, we establish power-efficiency tradeoffs and position the polytropic index as a genuine thermodynamic control knob for the rational design of high-speed, high-performance microscopic thermal machines.

In quantum-kinetic approaches to the dynamics of Coulomb-bound many-body correlations such as excitons, trions, biexcitons or higher-order correlations, a detailed knowledge of the many-body Coulomb Hamiltonian serving as a starting point is important. In this manuscript, the second-quantized description of the Coulomb interaction between Bloch electrons in a Heisenberg-equation-of-motion approach in atomically thin semiconductors is derived and reviewed. Emphasis is put on a discussion of Umklapp processes and the dielectric screening including all local-field effects. A link between \textit{ab initio} methods of screening and few-band models in effective-mass approximations for the quantum kinetics is established and all important Coulomb scattering processes contributing to the exciton energy landscape and density-independent exciton scattering are discussed.

Unidirectional magnetoresistance, or electric magnetochiral anisotropy (eMChA), is a nonlinear magnetotransport phenomenon that arises in noncentrosymmetric conductors , where changes in resistance $R(B)$ are: (i) chiral, $ΔR(B)/R(0)=2\,χ\, {\bf I}\cdot{\bf B}$, or (ii) polar, $ΔR(B)/R(0)=2\,γ\, {\bf I}\cdot({\bf P}\times{\bf B})$, with eMChA coefficients $χ$ and $γ$. In [Phys. Rev. Lett. 135, 106602 (2025)], we showed that the eMChA in the conduction band of tellurene is polar ($χ=0$, $γ\neq 0$) and emerges from the quantum metric dipole due to its Weyl node and from the lone pair polarization ${\bf P}$. Here, we extend our work to the valence band of tellurene, where the eMChA is usually said to be chiral ($χ\neq 0, γ= 0$). We show that also a polar coefficient $γ\neq 0$ emerges naturally through a downfolding procedure, in which remote Weyl-node containing bands induce momentum-space gradients of the quantum metric in the low-energy levels, activating finite metric dipoles. Combining semiclassical Boltzmann transport with a ${\bf k}\cdot{\bf p}$ description of tellurene, our numerical calculations agree quantitatively with doping ($μ$) dependent second-harmonic measurements of the longitudinal voltage $V^{2ω}_\parallel(μ)$ in perpendicular field. The combined chiral and polar characters ($χ\neq0, γ\neq 0)$ of the eMChA in tellurene also explains the shift in the angular ($φ$) dependence of $V^{2ω}_\parallel(φ)$ for in plane fields. Our results demonstrate that the polar eMChA can arise in topologically trivial bands through multiband effects and establishes tellurene as a platform for quantum-geometric rectification in both electron and hole regimes.

The increased complexity and reduced size of (opto-)electronic devices demands for quantitative descriptions of excess carrier transport and recombination via various mechanisms. In addition, experimental methods capable of resolving carrier dynamics on the nanometer scale are required. In this paper, we present a quantitative model of a confined geometry including recombination at two surfaces, which is very generic for electron beam induced current measurements in a scanning transmission electron microscope - a method which offers atomic scale spatial resolution. The model is based on analytical considerations as well as finite element simulations and underlying assumptions are subjected to an in-depth discussion. Finally, the successfull application to experimental data obtained on the complex oxide SrTi0.995Nb0.005O3 demonstrates the practicality and robustness of the approach, which enables the precise determination of its bulk diffusion length of L = 10.2 +- 0.1 nm.

Strong directional disorder in local magnetic moments coupled to a Chern insulator gives rise to topological phases that cannot be continuously connected to the clean limit and are therefore genuinely disorder-driven. We demonstrate this in a spinful Qi-Wu-Zhang model of a two-dimensional Chern insulator coupled to disordered classical spins of unit length. The topological phase diagram is computed numerically using two complementary approaches: twisted boundary conditions and the topological Hamiltonian technique. Our results show that strong disorder can act as a fundamental topological mechanism rather than merely a perturbation. For strong exchange coupling, tuning the mass parameter reveals a transition between phases with different Chern numbers $C$. Remarkably, this transition is driven by zeros, rather than poles, of the disorder-averaged Green's function crossing the chemical potential, and has no analogue in any clean system. We further identify a strong-coupling phase with $C = 0$ that is nonetheless topologically nontrivial, characterized by a distinct Chern number $C^{(\mathrm{S})} \neq 0$ over the manifold of classical spin configurations. This phase is also disorder-driven, as $C^{(\mathrm{S})} = 0$ in the clean limit.

In topological bands, it is impossible to construct exponentially localized Wannier functions while preserving the symmetries. Instead, in quantum Hall systems, one can define an overcomplete basis of spatially localized coherent states. In this work, we propose a unified framework for understanding the quantum Hall effect and Chern insulators from the perspective of localized bases, by extending the overcomplete basis of coherent states to Chern bands in terms of coherent-like states. Specifically, by representing both coherent states and coherent-like states as wave packets defined on a band, the difference between them can be encoded solely in the functional form of the wave packet in momentum space. Furthermore, for filling factor $ν=1/3$, we define a local repulsive interaction Hamiltonian based on these bases and discuss properties of its ground states. In particular, by relating this Hamiltonian to previously studied models, we show that in quantum Hall systems it possesses exactly zero-energy ground states with topological degeneracy, thereby confirming that it serves as a model for fractional quantum Hall systems. In addition, we numerically verify that the Hamiltonian possesses topological degeneracy for representative Chern insulator models. An advantage of this formulation is that it allows fractional quantum Hall systems and various fractional Chern insulator systems to be discussed within a unified framework using the same Hamiltonian form. In addition, we discuss that coherent-like states can also be defined in $\mathbb{Z}_2$ topological insulators. Corresponding to the fermionic time-reversal symmetry of the system, Kramers-degenerate coherent-like states can be naturally defined. The localized basis constructed from coherent-like states is expected to be useful for describing strongly correlated topological phases in flat-band systems.

What sustains a non-equilibrium system against fluctuations from within - as witnessed in non-equilibrium steady states, glassy relaxation, and even living organisms? Here we show that the arrow of time itself can be operationalized into a measurable physical quantity on mesoscopic particles - the eigen-phase displacement. This displacement gives rise to a non-local generalized force, the thermodynamic inertia force, which emerges from the integrated contribution of local constraints rather than as a conventional local force. It actively counteracts fluctuations and its algebraic structure is a semi-group, fundamentally distinct from the Lie group of Newton inertia, thereby encoding the irreversibility of time's arrow. Building on this foundation, we construct a unified Microstate-Sequence-Mode-Coupling (MSS-MCT) theory. Its thermodynamic limit is defined by Microstate Sequence (MSS) theory, and its dynamical action is captured by a consequent mode-coupling theory (MCT). From this single first-principles framework, we simultaneously resolve two long-standing puzzles: it predicts the giant non-Gaussian parameter(1~10), closing the order-of-magnitude gap with experiments that standard mode-coupling theory could not explain; and it delivers a first-principles, non-fitting derivation of the universal polymer constant $C_{1} \approx 16.7$ with merely 1 percent error - the most accurate theoretical prediction to date, dramatically surpassing the Adam-Gibbs and others. Our work establishes thermodynamic inertia as a foundational principle for non-equilibrium matter, bridging thermodynamic and dynamic descriptions from glassy relaxation to the maintenance of life.

We derive the Ward-Takahashi identity and establish the gauge-invariant response theory for open quantum systems described by Lindbladians to show that particle-number conservation is not necessary to satisfy gauge invariance. We construct an observable which can be used to test the gauge invariance in the absence of particle-number conservation. We derive the low-energy collective modes that emerge as a consequence of gauge invariance in open quantum systems, and find that two-body loss induces diffusive modes in dissipative Bardeen-Cooper-Schrieffer (BCS) superconductivity. Possible experimental situations for testing gauge invariance in open quantum systems are also discussed.

In ballistic transport, the movement of charged carriers is essentially unimpeded by scattering events. In this limit, microscopic parameters such as crystal momentum, spin and quantum phases are well conserved, allowing electrons to maintain their quantum coherence over longer distances. Nanoscale materials, like carbon nanotubes, graphene, and nanowires, exhibit ballistic transport. However, their scalability in devices is significantly limited. While deposited metal films offer excellent scalability for nanodevices, achieving ballistic transport in these films poses a challenge due to their short electronic mean free path. Here, we investigated the electronic transport in cross-geometry devices fabricated with 90 nm-thick copper films without grain boundaries. We observed ballistic transport in devices with channel width smaller than 150 nm below 85 K by measuring negative bend resistance. Our findings would open the opportunity for probing intrinsic quantum properties of Cu, and for realizing scalable low-loss signal transmission and high-quality interconnects in semiconductor devices.

Spontaneous dynamic systems have attracted significant attention for their rich underlying physics such as phase-locking and synchronization. In this work, we report a new mechanism of generating magnetic spontaneous oscillation via parametric pumping. By applying a pump tone to excite propagating spin waves in a yttrium iron garnet delay line, four-wave mixing converts the pump mode into two phase-autonomous propagating magnon modes, i.e. a spontaneous mode with nearly twice the wavenumber of the pump mode and an idler mode with nearly zero wavenumber. This allows us to reliably generate ultrasharp spin wave dynamics with broad frequency tunability from the pump and magnetic field. We show that the spontaneous mode can be phase-locked to a probe tone, similar to an auto-oscillator. Furthermore, the spontaneous dynamics can be used to implement a high-gain magnonic parametric amplifier with a gain up to 40 dB. Our results open a new avenue for studying nonlinear magnonics and synchronization physics in propagating magnon geometry and for developing new magnonic devices.

Stacking geometry in multilayer graphene (MLG) provides an interesting degree of freedom to engineer its electronic structure near the Fermi level, wherein the linear bands in single layer graphene could retain or evolve into parabolic or flat bands. Using a tight-binding model, we carried out a detailed analytical analysis of the electronic band structures for arbitrarily stacked MLGs. We show that their low energy band dispersions near the Fermi level may be deduced from its substacks in isolation. The analytical solutions of the momenta with zero eigenvalue for an AA stacking allows us to generalize the results of the zero energy momenta for arbitrarily stacked MLGs. Moreover, we find that an interplay of parallel and rhombohedral stackings allows for flat band engineering and enhancement in arbitrarily stacked MLGs. The existence of flat bands in MLGs might offer another interesting platform for exploring the superconductivity in graphene systems beyond the twisted bilayer graphene.

We theoretically propose and analyze a Josephson-like magnetic tunnel junction (MTJ) structure that exhibits quantum spin dynamics analogous to those in superconducting Josephson junctions. By exploiting the isomorphism between the equations of motion for low-dissipation MTJs with easy-plane anisotropy and the Josephson phase dynamics, we construct a theoretical framework for realizing spintronic qubits. Within this framework, we identify the physical parameters -- such as anisotropy constants, Gilbert damping, spin current amplitude, and geometric factors -- that govern the transition from classical to quantum behavior. We show that different types of spintronic qubits, including analogs of charge, flux, and transmon superconducting qubits, can be implemented depending on the hierarchy of energy scales. A Hamiltonian formalism is developed for each regime, enabling an analytical treatment of the two-level quantum dynamics and estimation of coherence times. In particular, we demonstrate that the spin current can be used not only to excite but also to stabilize the qubit states through dissipation control. These findings provide a route toward integrating spintronic qubits into CMOS-compatible architectures and lay the groundwork for a fully spintronic platform for quantum computation.

This letter presents a model system for controllable two-magnon interference in the time domain. This two-magnon interference, i.e., a magnonic analog to the photonic Hong-Ou-Mandel effect, is supported by a tunable magnonic beamsplitter operation formed in a hybrid cavity magnonic system comprising a pair of mutually coupled magnon modes. By applying a time-dependent magnetic field, magnons can be excited independently in each mode and subsequently brought into interaction, shifting from independent to collective oscillations, to realize a controllable magnonic beamsplitter. When the beamsplitter operation is applied to an initially unentangled two-magnon state, a maximally entangled magnonic $N00N$ state with tunable phase sensitivity is produced. These findings suggest that two-magnon interference in hybrid cavity magnonic systems may enable novel quantum metrological devices to study fundamental magnon dynamics and contribute to developing hybrid magnonic quantum computing architectures.

Previous work of Ayyer, Martin, and Williams gave a probabilistic interpretation of the Macdonald polynomials $P_λ(x_1,\dots,x_n;1,t)$ at $q=1$ in terms of a Markov chain called the multispecies $t$-Push TASEP, a Markov chain involving particles of types $λ_1,\dots,λ_n$ hopping around a ring. In particular, they showed that for each composition $η$ obtained by permuting the parts of $λ$, the stationary probability of being in state $η$ is proportional to the ASEP polynomial $F_η(x_1,\dots,x_n; 1,t)$, and the normalizing constant (or partition function) is $P_λ(x_1,\dots,x_n; 1,t)$. There is an inhomogeneous generalization of Macdonald polynomials due to Knop and Sahi called interpolation Macdonald polynomials $P^*_λ(x_1,\dots,x_n;q,t)$, as well as an inhomogeneous generalization of ASEP polynomials called interpolation ASEP polynomials $F^*_η(x_1,\dots,x_n;q,t)$ that we introduced in previous work. In this article we introduce a new Markov chain called the interpolation $t$-Push TASEP, and show that its steady state probabilities and partition function are given by the interpolation ASEP polynomials and the interpolation Macdonald polynomial, evaluated at $q=1$. This generalizes the previous result of Ayyer, Martin, and Williams.

Charged excitons, or trions, offering unique spin and charge degrees of freedom, have primarily been investigated in doped systems where charges are long considered indispensable. Here, we present an alternative route to ultra-efficient trion emission from an intrinsic, defect-free semiconductor via a transfer mechanism. By exciting trions in two-dimensional tungsten-diselenide donors and transferring them into one-dimensional carbon-nanotube acceptors in mixed-dimensional heterostructures, we circumvent the usual carrier requirement, overcoming intrinsic Auger-quenching limitations. Benefitting from a reservoir effect induced by dimensional heterogeneity, this process achieves trion emission efficiencies increased by over 100-fold compared to conventional doping-based approaches, and remains robust across diverse doping conditions. Our findings extend the exciton transfer paradigm to the three-body quasiparticles, offering a new platform for advancing excitonic physics and trion-based optoelectronic/spintronic applications.

Dense amorphous materials exhibit both nonlocal flow cooperativity and pronounced history dependence, yet existing continuum models capture only one of these features at a time. Nonlocal rheologies are intrinsically memoryless, while thixotropic models remain local. Here we introduce a coupling between structural memory and nonlocal fluidity to include aging and rejuvenation in nonlocal granular fluidity. The resulting model reproduces hysteresis in shear-rate sweeps and delayed yielding in creep, while preserving nonlocal flow profiles. By introducing memory augmented non local granular fluidity, MNGF, we show that nonlocality alone cannot encode history, and memory alone cannot encode spatial cooperativity, but their coupling is essential and minimal. These results demonstrate that memory and nonlocality must be treated jointly to describe history dependent flows, and provide a unified framework for modeling time-dependent rheology in dense amorphous materials.

Classical ecological models predict that large, diverse communities should be unstable, presenting a central challenge to explaining the stable biodiversity seen in nature. We revisit this long-standing problem by extending the generalized Lotka-Volterra model to include both spatial structure and environmental fluctuations across space and time. We find that neither space nor environmental noise alone can resolve the tension between diversity and stability, but that their combined effects permit arbitrarily many species to stably coexist despite strongly disordered competitive interactions. We analytically characterize the noise-induced transition to coexistence, showing that spatiotemporal noise drives an anomalous scaling of abundance fluctuations, known empirically as Taylor's law. At the community level, this manifests as an effective sublinear self-inhibition that renders the community stable and asymptotically neutral in the high-diversity limit. Spatiotemporal noise thus provides a novel resolution to the diversity-stability paradox and a generic mechanism by which complex communities can persist.

Metastability, i.e., partial relaxation to long-lived, quasi-stationary states before true asymptotic equilibrium sets in, emerges ubiquitously in classical and quantum dynamical systems as a result of timescales separation. In open quantum systems, an intrinsically nonequilibrium analogue, dynamical metastability, can originate from the spectral geometry of a non-Hermitian operator. In noninteracting models, this mechanism produces boundary-sensitive anomalous relaxation, transient amplification, and topologically mandated long-lived edge modes, all of which are enhanced as system size grows. Here we extend dynamical metastability into the nonlinear, interacting regime and identify magnetic heterostructures as a natural platform for its exploration. We introduce an interacting spin Lindbladian whose linearized magnon dynamics map onto a dynamically metastable Hatano-Nelson chain, and show that dynamical metastability in the noninteracting limit seeds genuinely nonlinear phenomena, including size-dependent spin dipping and anomalous attraction to unstable equilibria. Long-lived edge states associated to topologically mandated Dirac bosons persist under nonlinearities and disorder. We further analyze the magnetization dynamics in magnetic multilayers within the classical Landau-Lifshitz-Gilbert-Slonczewski framework, identifying Dzyaloshinskii-Moriya interaction, nonlocal damping, and spin-transfer torque as control parameters governing bulk-boundary stability mismatch and band topology. While all the distinctive dynamical phenomena previously identified reappear in this experimentally relevant setting, the LLGS framework also supports multistability and limit cycles that are absent in the quantum model. Our results constitute the first systematic study of dynamical metastability in nonlinear dynamics, directly relevant to spin-torque oscillator arrays, magnonic devices, and beyond.

Entanglement is the hallmark of quantum physics, yet its characterization in interacting many-body systems at thermal equilibrium remains one of the most important challenges in quantum statistical physics. We prove that the Gibbs state of any quantum spin chain can be exactly decomposed into a mixture of matrix product states with a bond dimension that is independent of the system size, at any finite temperature. As a consequence, the Schmidt number, arguably the most stringent measure of bipartite entanglement, is strictly finite for thermal states, even in the thermodynamic limit. Our decomposition is explicit and is accompanied by an efficient classical algorithm to sample the resulting matrix product states.

In this study, we demonstrate velocity-dependent directional friction on a surface structured with tilted (~57°) titania nanorods using standard and colloidal probe force microscopy. Friction is measured at four different sliding speeds in two configurations, along and opposite to the tilt and perpendicular to the tilt direction, exhibiting anisotropic friction. Furthermore, friction decreases logarithmically with increasing sliding speed, which is attributed to the viscoelastic bending of the nanorods caused by stress-induced defect migration. The velocity weakening is more pronounced in the direction perpendicular to the tilt than along and opposite to it. The experimental findings are corroborated by creep measurements, which are well-reproduced by the Standard Linear Solid (SLS) model of viscoelasticity. Our results may be applied to the development of direction- and velocity-dependent sensors for microscale sliding motion as a robust alternative to structured interfaces based on polymeric materials.

Superparamagnetism in nanoscale magnetic layers is a critical property for a wide range of spintronic-based sensor and computing applications. While conventional magnetization measurements can detect superparamagnetic signatures, they often require the application of high perturbative fields and are difficult to implement for magnetic layers integrated within functional devices. In this study, we non-invasively investigate the superparamagnetic spin dynamics of a nanoscale CoFeB layer of thickness 1.1 nm, deposited on a diamond substrate, by probing its low-frequency near-field electromagnetic (EM) fluctuations using nitrogen-vacancy (NV) centers-based quantum dephasometry. Our measurements reveal an unconventional, non-monotonic temperature dependence of the dephasing time of NV centers, which we attribute to EM fluctuations produced by thermally driven superparamagnetic domain flipping in CoFeB. Our findings are further supported by the theoretical interpretation of the dephasing dynamics of NV centers and the complementary SQUID-based magnetization characterizations of the CoFeB layer. Additionally, exploiting the technique of NV dephasometry, we extract the spectral density of the EM fluctuations in CoFeB, which is used to isolate different components of the EM fluctuations acting on NV centers. We also measure the CoFeB-to-NV distance-dependent coherence times of NV centers to investigate the effect of the dimensionality of the CoFeB layer on the generated near-field EM fluctuations. These results provide critical insight into the magnetization dynamics and near-field EM environment of nanoscale magnetic layers. It also has significant implications for the development of hybrid quantum spintronic devices and applications involving nanoscale opto-magnetic materials.

In this letter we discuss the classical integrable elliptic Toda chain proposed by I. Krichever. Our goal is to construct an open elliptic Toda chain with boundary terms. This is achieved using the factorized form of the Lax matrix and gauge equivalence with the XYZ chain.

The extraction of spatio-temporal coherence in high-dimensional, chaotic, non-linear dynamical systems, such as turbulent flows, remains a fundamental challenge in physics, mathematics and engineering. In this work, we employ Shannon transfer entropy (TE) to identify causally coherent motions in a zero-pressure-gradient turbulent boundary layer (TBL). This causality metric, rooted in information theory, enables the identification of sources and targets in dynamical systems using the corresponding time series. However, TE requires sophisticated tuning of various hyperparameters, such as the Markovian order of the source ($m$), which can spatially vary in wall-bounded turbulent flow. Here, we present an adaptive tuning and discuss the influence of $m$ across different TBLs. We introduce the concept of causally coherent structures (CCS), i.e. coherent structures interpreted as spatio-temporal patterns of causality. Moreover, the net transfer entropy flux is also utilised to identify boundary layer locations acting either as sources or targets. The standard viscous, logarithmic, and outer layers are characterised by information fluxes, highlighting, for example, dominant top-down interactions between the inner and outer layers, analogously to the classical energy cascade. This work extends techniques previously employed in the literature, such as correlation and spectral analysis, and presents an approach that is inherently general and applicable to a wide range of chaotic dynamical systems, with applications in cognitive sciences, systems biology and finance.

We study a probabilistic cellular automaton obtained as a mixture of the additive elementary rules 60 and 102. We prove that, for any finite periodic lattice and for mixing parameter $λ=1/2$, the system almost surely reaches the absorbing all-zero configuration in finitely many steps. In addition, Monte Carlo simulations indicate as well the presence of a zero-density stationary state in a finite interval around $λ=1/2$. Despite this absorbing behavior, both mean-field and block approximation schemes predict a stationary state with non-zero density. This failure, traced to the additive and mirror symmetries of the deterministic components, highlights a fundamental limitation of finite-block approximation in capturing the global dynamics of probabilistic cellular automata.

The spikes train is an important step in order to the artificial neural network (ANN) give us simulations more close to the reality i.e the operation of the biological neural network. Based on in previous our work that the HANN can to produce critical and tricritical intermittencies we investigate in present work the possibility of the Spike train production from the HANN. So the operation of ANN does not would based in mathematical algorithm of machine learning but the operation of a ANN could be based in physical notions as the phenomenon of intermittency. As we have shown the real biological neurons is a Dynamical system which present the intermittent dynamic type I.

In this paper, we consider the Cauchy problem for the defocusing nonlinear Schr$\ddot{\text{o}}$dinger equation with a finite genus algebro-geometric background. Long-time asymptotics of the solution are derived in four space-time regions. It comes out that the leading-order term in the expansion is, up to a constant, given by the background solution with a shift of the parameter. The subleading term, however, decays at different rates for different regions. We particularly highlight that in the two transition regions, they are of order $\mathcal{O}(t^{-1/3})$ and the coefficients involve an integral of the Painlevé XXXIV transcendent. We establish our results by applying a nonlinear steepest descent analysis to the associated Riemann-Hilbert problems.

In Japan, severe rice shortages in 2024 sparked widespread public controversy across both news media and social platforms, culminating in what has been termed the "Reiwa Rice Riot." This study proposes a framework to analyze the temporal dynamics and causal interactions of emotions expressed on X (formerly Twitter) and in news articles, using the "Reiwa Rice Riot" as a case study. While recent studies have shown that emotions mutually influence each other between social and mass media, the patterns and transmission pathways of such emotional shifts remain insufficiently understood. To address this gap, we applied a machine learning-based emotion classification grounded in Plutchik's eight basic emotions to analyze posts from X and domestic news articles. Our findings reveal that emotional shifts and information dissemination on X preceded those in news media. Furthermore, in both media platforms, the fear was initially the most dominant emotion, but over time intersected with hope which ultimately became the prevailing emotion. Our findings suggest that patterns in emotional expressions on social media may serve as a lens for exploring broader social dynamics.

In this paper, we introduce a gauge-mediated Epidemiological Model inspired by Quantum Electrodynamics (QED). In this model, the ``direct contact'' paradigm of classical SIR models is replaced by a gauge-mediated interaction where the environment, represented by a pathogen field $\varphi$, plays a fundamental role in the epidemic dynamics. In this model, the non-local characteristics of epidemics appear naturally by integrating out the pathogen field. Utilizing the Doi-Peliti formalism, we derive the effective action of the system and the standard Feynman rules that can be used to compute perturbatively any observables. Using standard QED techniques, we show how to relate renormalized pathogen mass, Debye screening, to epidemiological concepts and we compute at first order the effective reproductive number,$R_{eff}$, and how the condition to have an epidemic is related to a phase transition in the pathogen mass. We show that the superspreading hosts can be included easily in this formalism.

The design of high-performance optical systems remains a specialized domain gated by the limited availability of expert engineers, creating a bottleneck that stalls innovation despite the growing demand for imaging hardware. While deep learning has improved parameter optimization, it has yet to address the fundamental challenge of conceptualizing valid optical architectures from functional requirements. Here, we present an end-to-end generative framework that couples the semantic reasoning of Large Language Models (LLMs) with a differentiable ray-tracing engine to democratize the synthesis of diffraction-limited optical prescriptions. By treating optical design as a semantic-to-physical translation task, the system autonomously interprets prompts ranging from high-level end-user requests to rigorous technical specifications. We demonstrate the framework's versatility across three distinct regimes: (1) finite-conjugate industrial metrology systems, where the model autonomously enforces application-specific constraints such as telecentricity to achieve diffraction-limited performance; (2) a suite of infrared objectives (NIR, SWIR, and LWIR), demonstrating the framework's ability to synthesize valid topologies and optical prescriptions for non-visible spectral bands, and (3) complex aspheric mobile lenses, where the system successfully navigates the high-dimensional optimization landscape to produce high-resolution designs suitable for modern sensors. Validated against industry-standard simulation tools, these results establish a new paradigm for automated optical engineering, bridging the gap between semantic intent and physical realization.

Synthetic survey data generated by large language models (LLMs) suffers from a fundamental circularity: the same model family that generates text responses also maps them to numerical scales. We calibrate and validate Semantic Similarity Rating (SSR; Maier et al., 2024), which decouples generation from scale mapping via embedding-based cosine similarity against predefined anchor statements. Configuration experiments (N=17 pilot, N=69 cross-validation across 8 domains) show that naturalistic behavioral anchors outperform formal jargon by 29 percentage points (pp), and that SSR achieves 65-67% exact match and 91% within plus/minus 1; a cross-model test with OpenAI text-embedding-3-small reaches 77% exact, confirming cross-provider generalization. Direct LLM baselines (Claude 87%, GPT-4o 83%) establish that SSR's contribution is methodological independence, not accuracy superiority. A control condition removing question text from the LLM prompt actually improves LLM accuracy, ruling out information asymmetry as the explanation for SSR's lower accuracy. A pre-registered circularity experiment (N=345) reveals 4x compressed error variance in LLM rating (sigma^2 = 0.21 vs 0.87 for SSR) and systematic directional bias. A cross-model control (GPT-4o rating Claude-generated text) shows nearly identical compression (within/cross ratio = 0.93), indicating variance compression is a general LLM property rather than a within-model artifact. The calibration dataset, anchor library, and source code are publicly available (see Data Availability).

We present an anisotropic goal-oriented error estimator based on the Dual Weighted Residual (DWR) method for time-dependent convection-dominated problems. Using elementwise p-anisotropic finite element spaces, the estimator is elementwise separated with respect to the single directions in space and time. This naturally leads to adaptive, anisotropic hp-refinement (h-anisotropic refinement and elementwise anisotropic p-enrichment). We employ discontinuous elements in space and time, which are well suited for problems with high Peclet numbers. Efficiency and robustness of the underlying algorithm are demonstrated for different goal functionals. The directional error indicators quantify anisotropy of the solution with respect to the goal, and produce hp-refinements that efficiently capture sharp layers. Numerical examples in up to three spatial dimensions demonstrate the superior performance of the proposed method compared to isotropic h and hp adaptive refinement using established benchmarks for convection-dominated transport.

Innovation records often exhibit "hockey-stick" patterns of abrupt, near-singular growth at the collective level. However, this macroscopic explosiveness stands in stark contrast to individual discovery, which remains bounded by cognitive and temporal constraints and follows slow, sublinear accumulation laws. Here, we resolve this micro-macro discrepancy by introducing a minimal multi-scale model that identifies the growth of the explorer population as the primary driver of aggregate acceleration. Building on the Theory of the Adjacent Possible and the Urn Model with Triggering (UMT), we demonstrate that as discoveries expand the space of possibilities, they attract new explorers through a self-reinforcing branching process. This expansion induces a nonlinear mapping between intrinsic time (individual discovery events) and natural time (calendar years), effectively reparameterizing steady individual trajectories into accelerating system-level dynamics. We validate the framework using large-scale patent (EPO) and scientific publication (OpenAlex) datasets, showing that the model accurately reproduces stable per-capita productivity alongside exponential aggregate growth. By providing a quantitative link between individual behavior and collective takeoffs, this work offers a unified foundation for understanding the statistical structure and temporal evolution of innovation ecosystems.

We develop and analyze a model for a flat microbial droplet growing on the surface of a three-dimensional viscous fluid. The model describes growth-induced stresses at the fluid surface, density variations in the bulk due to nutrient consumption, and the resulting fluid flows that arise. We reformulate this free-boundary problem as a system of integro-differential equations defined solely on the microbial domain. From this formulation, we identify an axisymmetric solution corresponding to a radially expanding disk and analyze its morphological stability. We find that growth forces stabilize the axisymmetric solution while buoyancy forces destabilize it. We connect these findings to experimental observations.

Collective decision-making can be viewed as the problem of aggregating multiple noisy information channels about an unknown state of the world. Classical epistemic justifications of majority rule rely on restrictive assumptions about the homogeneity and symmetry of these channels, which are often violated in realistic environments. This paper introduces the Epistemic Shared-Choice Mechanism (ESCM), a lightweight and auditable procedure that endogenously estimates issue-specific signal reliability and assigns bounded, decision-specific voting weights. Using central limit approximations, the paper provides an analytical comparison between ESCM and unweighted majority rule, showing how their relative epistemic performance depends on the distributional structure of information in the population, including unimodal competence distributions and segmented environments with informed minorities. The results indicate that endogenous and bounded epistemic weighting can improve collective accuracy by merging procedural and epistemic requirements.

We present a quantum algorithmic framework for simulating phase-contrast transmission electron microscopy (CTEM) image formation using a fault-tolerant, gate-based quantum circuit model. The electron wavefield on an $N\times N$ grid is amplitude-encoded into a $2\log_2 N$-qubit register. Free-space propagation and objective-lens aberrations are implemented via two-dimensional quantum Fourier transforms (QFTs) and diagonal phase operators in reciprocal space, while specimen interaction is modeled under the weak phase object approximation (WPOA) as a position-dependent phase grating. We validate projected potentials, contrast transfer function (CTF) behavior, and image contrast trends against classical multislice simulations for MoS$_2$ over experimentally relevant parameters, and provide resource estimates and key assumptions that determine end-to-end runtime. While extracting complete $N\times N$ intensity images requires $O(N^2/ε^2)$ measurements that preclude advantage for full-image reconstruction, the framework enables quantum advantage for tasks requiring Fourier-space queries, global image statistics, or phase-coherent observables inaccessible to classical intensity-only detection. This framework provides a physics-grounded mapping from CTEM theory to quantum circuits and establishes a baseline for extending toward full multislice and inelastic scattering models.

Accurate many-body treatments of condensed-phase systems are challenging because correlated solvers such as full configuration interaction (FCI) and the density matrix renormalization group (DMRG) scale exponentially with system size. Downfolding and embedding approaches mitigate this cost but typically require prior selection of a correlated subspace, which can be difficult to determine in heterogeneous or extended systems. Here, we introduce a stochastic cluster expansion framework for efficiently recovering the total correlation energy of large systems with near-DMRG accuracy, without the need to select an active space a priori. By combining correlation contributions from randomly sampled environment orbitals with an exactly treated subspace of interest, the method reproduces total energies for non-reacting and reactive systems while drastically reducing computational cost. The approach also provides a quantitative diagnostic for molecule-solvent correlation, guiding principled embedding decisions. This framework enables systematically improvable many-body calculations in extended systems, opening the door to high-accuracy studies of chemical processes in condensed phase environments.

Recurrent neural networks (RNNs) provide a theoretical framework for understanding computation in biological neural circuits, yet classical results, such as Hopfield's model of associative memory, rely on symmetric connectivity that restricts network dynamics to gradient-like flows. In contrast, biological networks support rich time-dependent behaviour facilitated by their asymmetry. Here we introduce a general framework, which we term drift-diffusion matching, for training continuous-time RNNs to represent arbitrary stochastic dynamical systems within a low-dimensional latent subspace. Allowing asymmetric connectivity, we show that RNNs can faithfully embed the drift and diffusion of a given stochastic differential equation, including nonlinear and nonequilibrium dynamics such as chaotic attractors. As an application, we construct RNN realisations of stochastic systems that transiently explore various attractors through both input-driven switching and autonomous transitions driven by nonequilibrium currents, which we interpret as models of associative and sequential (episodic) memory. To elucidate how these dynamics are encoded in the network, we introduce decompositions of the RNN based on its asymmetric connectivity and its time-irreversibility. Our results extend attractor neural network theory beyond equilibrium, showing that asymmetric neural populations can implement a broad class of dynamical computations within low-dimensional manifolds, unifying ideas from associative memory, nonequilibrium statistical mechanics, and neural computation.

When mutations are rampant, quasispecies theory or Eigen's model predicts that the fittest type in a population may not dominate. Beyond a critical mutation rate, the population may even be delocalized completely from the peak of the fitness landscape and the fittest is ironically lost. Extensive efforts have been made to understand this exceptional scenario. But in general, there is no simple prescription that predicts the eventual degree of localization for arbitrary fitness landscapes and mutation rates. Here, we derive a simple and general relation linking the quasispecies' Hill numbers, which are diversity metrics in ecology, and the ratio of an effective fitness variance to the mean mutation rate squared. This ratio, which we call the localization factor, emerges from mean approximations of decomposed surprisal or stochastic entropy change rates. On the side of application, the relation we obtained here defines a combination of Hill numbers that may complement other complexity or diversity measures for real viral quasispecies. Its advantage being that there is an underlying biological interpretation under Eigen's model.

Social entities only exist in virtue of collective acceptance or recognition, or acknowledgement by two or more individuals in the context of joint activities. Joint activities are made possible by the coordination of plans for action, and the coordination of plans for action is made possible by the capacity for collective intentionality. This paper investigates how primitive is the capacity that nonhuman animals have to create social entities, by individuating how primitive is the capacity for collective intentionality. I present a novel argument for the evolutionary primitiveness of social entities, by showing that the collective intentions upon which these social entities are created and shared are metaphysically reducible to the relevant individual intentions.

Effective representations of protein sequences are widely recognized as a cornerstone of machine learning-based protein design. Yet, protein bioengineering poses unique challenges for sequence representation, as experimental datasets typically feature few mutations, which are either sparsely distributed across the entire sequence or densely concentrated within localized regions. This limits the ability of sequence-level representations to extract functionally meaningful signals. In addition, comprehensive comparative studies remain scarce, despite their crucial role in clarifying which representations best encode relevant information and ultimately support superior predictive performance. In this study, we systematically evaluate multiple ProtBERT and ESM2 embedding variants as sequence representations, using the adeno-associated virus capsid as a case study and prototypical example of bioengineering, where functional optimization is targeted through highly localized sequence variation within an otherwise large protein. Our results reveal that, prior to fine-tuning, amino acid-level embeddings outperform sequence-level representations in supervised predictive tasks, whereas the latter tend to be more effective in unsupervised settings. However, optimal performance is only achieved when embeddings are fine-tuned with task-specific labels, with sequence-level representations providing the best performance. Moreover, our findings indicate that the extent of sequence variation required to produce notable shifts in sequence representations exceeds what is typically explored in bioengineering studies, showing the need for fine-tuning in datasets characterized by sparse or highly localized mutations.

We previously investigated color constancy in photorealistic virtual reality (VR) and developed a Deep Neural Network (DNN) that predicts reflectance from rendered images. Here, we combine both approaches to compare and study a model and human performance with respect to established color constancy mechanisms: local surround, maximum flux and spatial mean. Rather than evaluating the model against physical ground truth, model performance was assessed using the same achromatic object selection task employed in the human experiments. The model, a ResNet based U-Net from our previous work, was pre-trained on rendered images to predict surface reflectance. We then applied transfer learning, fine-tuning only the network's decoder on images from the baseline VR condition. To parallel the human experiment, the model's output was used to perform the same achromatic object selection task across all conditions. Results show a strong correspondence between the model and human behavior. Both achieved high constancy under baseline conditions and showed similar, condition-dependent performance declines when the local surround or spatial mean color cues were removed.

In the context of multistability driven diseases, like cancer, spatiotemporal plasticity plays a significant role to achieve a spectrum of phenotypic variations. The interplay between gene regulatory networks and environmental factors, such as resource competition and spatial diffusion, plays a crucial role in determining cellular behaviour and phenotypic heterogeneity. Though reaction diffusion frameworks have been widely applied in developmental biology, less attention has been paid to the simultaneous effects of resource competition and growth feedback on spatial organization. In this paper, we observed that a bistable genetic circuit under high resource competition due to growth feedback gives rise to multiple emergent phenotypes, as observed in cancer systems. Furthermore, we observed how spatial diffusion coupled with intrinsic nonlinearity can drive the emergence of distinct spatial dynamics over time. The observed spatiotemporal plasticity can also be driven by the comparative stability of the fixed points, diffusivity, and asymmetry of diffusion. Our findings highlight that growth-induced resource competition combined with diffusion can provide deeper insights into metastasis and cancer progression.

This study investigates the relationship between interocular distance relative to overall facial width (width ratio) and perceived subjective beauty in male faces. Building on the methodology of Pallett et al. (2010), who found that average proportions in female faces were rated as most attractive, the current study aimed to test this hypothesis in male faces. Faces from the Chicago Face Database (Ma et al., 2015) were morphed into average faces within three groups (with low, medium, and high width ratios), each composed of 96 or 97 individual images. These three average faces were then systematically manipulated in their width ratios across three levels in both directions, respectively, resulting in a total of 21 comparable faces. The use of multiple base faces served as a control for potential artifacts of image processing. Consequently, comparisons were restricted to within-group pairs to avoid confounding by co-varying facial features (e.g., skin tone), which precluded direct cross-condition comparisons but ensured internal validity. In a two-alternative forced-choice task, participants selected the more beautiful face from each pair. The data were analyzed using a Bayesian model which enables inference of the width ratio perceived as most beautiful. Results support the hypothesis that averageness in facial proportions correlates with higher perceived attractiveness. The study highlights the importance of controlling for image manipulation, including attempts at methodological implementation, and of considering ethnicity as a potential moderating variable. These findings offer a data-driven foundation for understanding facial aesthetics and cognitive processes of human perception, with applications in advertising, artificial face generation, and plastic surgery.

Recent experiments in moiré quantum materials exhibit quantized Hall states without an external magnetic field, motivating continuum mechanisms based on smooth moiré-periodic pseudospin textures. We present a controlled theory of skyrmion Chern bands generated by such textures. An exact local $SU(2)$ transformation reveals an emergent non-Abelian gauge field; for large branch splitting we perform an operator-level Schrieffer-Wolff expansion, yielding a single-branch Hamiltonian together with systematically dressed physical operators that define the projected interacting theory beyond strict adiabaticity. The leading dynamics is governed by a $U(1)$ Berry connection whose flux is set by the skyrmion density, while controlled non-adiabatic corrections are fixed by the texture's real-space quantum geometric tensor. In a Landau-level representation built from the averaged emergent field, moiré-periodic modulations induce Umklapp-resolved deformations of Girvin-MacDonald-Platzman kinematics and microscopic sources of excess optical quantum weight above the topological lower bound. Assuming a gapped Hall phase, we further derive a skyrmion-crystal effective field theory with a universal Berry-phase term and a noncommutative magnetophonon. Our results provide experimentally accessible signatures for twisted transition-metal dichalcogenide homobilayers and rhombohedral graphene aligned with hexagonal boron nitride.

The unitary coupled cluster (UCC) algorithm is one of the most promising implementations of the variational quantum eigensolver for quantum computers. However, for large systems, the number of UCC factors leads to deep quantum circuits, which are prohibitive for execution on quantum hardware. To address this, circuit depth can be reduced at the cost of more measurements with a Taylor series expansion of UCC factors with small angles, while treating the large-angle factors exactly. We implement this approach to quadratic order (qUCC) for systems with strong correlations and systems where conventional methods like coupled cluster (CC) with low excitation levels fail, but UCC and qUCC perform well. We study hydrogen chains and the BeH2 molecule that allow us to change the degree of strong correlation due to geometrical distortions. We show, via a dramatic increase in number of factors able to handle exactly, a systematic convergence of these results as more exact UCC factors are included in the calculations -- the hardest to converge regime is in the crossover from weak to strong coupling. In all cases the total number of UCC factors needed to be treated exactly is much less than the total number of UCC factors available (typically about one-third to one-half of the total number of factors).

Programmability is increasingly central in emerging quantum network software stacks, yet the node-internal controller-to-hardware interface for quantum repeater devices remains under-specified. We introduce the idea of an instruction-set architecture (ISA) for controller-driven programmability of nitrogen-vacancy (NV) center quantum repeater nodes. Each node consists of an optically interfaced electron spin acting as a data qubit and a long-lived nuclear-spin register acting as a control program. We formalize two modes of programmability: (i) deterministic register control, where the nuclear register is initialized in a basis state to select a specific operation on the data qubit; and (ii) coherent register control, where the register is prepared in superposition, enabling coherent combinations of operations beyond classical programmability. Network protocols are expressed as controller-issued instruction vectors, which we illustrate through a compact realization of the BBPSSW purification protocol. We further show that coherent register control enables interferometric diagnostics such as fidelity witnessing and calibration, providing tools unavailable in classical programmability. Finally, we discuss scalability to multi-electron and multi-nuclear spin architectures and connection to Linear combination of unitaries (LCU) and Kraus formulation.

A fascinating effect belonging to the field of vacuum forces and fluctuations is that of quantum friction. It refers to the prediction of a dissipative force acting on a moving object due to the quantum vacuum field. In this work, we investigate rotational quantum friction where a diatomic polar molecule rotates around its own center of mass in free space. We quantize the rotational motion and investigate the resulting dissipation due to spontaneous decay. We find in the Markovian regime that a friction torque $\propto Ω^3$ persists even for zero temperature, and in agreement with the classical result in the limit of large rotational quantum number $l$. Within the non-Markovian short-time regime we find a friction $\proptoΩ$.

The Quantum Approximate Optimization Algorithm (QAOA) is a leading approach for combinatorial optimization on near-term quantum devices, yet its scalability is limited by the difficulty of optimizing \(2p\) variational parameters for a large number \(p\) of layers. Recent empirical studies indicate that optimal QAOA angles exhibit concentration and transferability across problem sizes. Leveraging this observation, we propose a schedule-learning framework that transfers spectral-gap-informed adiabatic control strategies from small-scale instances to larger systems. Our method extracts the spectral gap profile of small problems and constructs a continuous schedule governed by \(\partial_t s = κg^q(s)\), where \(g(s)\) is the instantaneous gap and \((κ, q)\) are global hyperparameters. Discretizing this schedule yields closed-form expressions for all QAOA angles, reducing the classical optimization task from \(2p\) parameters to only \(2\), independent of circuit depth. This drastic parameter compression mitigates classical optimization overhead and reduces sensitivity to barren plateau phenomena. Numerical simulations on random QUBO and 3-regular MaxCut instances demonstrate that the learnt schedules transfer effectively to larger systems while achieving competitive approximation ratios. Our results suggest that gap-informed schedule transfers provide a scalable and parameter-efficient strategy for QAOA.

The article deals with Gödel-like solutions in the context of Galilean gravity, a geometric formulation of non-relativistic gravitation defined on a five-dimensional Galilean manifold. Within this framework, non-relativistic matter fields admit a covariant description, while the physical Newtonian dynamics is recovered through an immersion into the usual $3+1$ spacetime. By adopting a Gödel-like metric ansatz and coupling the gravitational field to a Galilean fluid derived from a variational principle, we obtain a system of highly nonlinear and coupled field equations. Exact solutions are constructed by fixing the matter sector consistently with the field equations. The resulting configurations describe rotating non-relativistic universes and satisfy $D(x)>H(x)$ throughout the entire spatial domain. As a consequence, the associated Killing vector remains spacelike everywhere and no closed timelike curves arise.

We argue that for finite energy windows, the final states in gravitational scattering in dimension $d > 4$ are normalizable coherent states in Fock space. However, as the center of the energy window goes to infinity, black hole physics predicts that these states become orthogonal to every state with a finite number of particles. Given that the spectral measure in energy is determined by Poincare invariance, the S-matrix cannot be a unitary operator in Fock space, despite having finite matrix elements in Fock space, and satisfying perturbative unitarity, to all orders in string perturbation theory. We identify regimes in the BFSS matrix model\cite{bfss} and the definition of the S-matrix as the limit of CFT correlators\cite{polchsuss}, which point to the same conclusion. We review a scattering theory based on the quantum mechanics of a finite number of fermionic oscillators, whose algebra formally converges to the Super-Poincare covariant Awada-Gibbons-Shaw\cite{ags} algebra, and argue that a certain class of limiting states on that algebra satisfy all the properties required by physical unitarity in the algebraic formulation of quantum mechanics. The only missing ingredient for a consistent theory is a proof that the S matrix amplitudes themselves are Poincare invariant. We provide suggestive arguments, but no real proof, that this is so.

Golden dark sirens - exceptionally well-localized gravitational-wave (GW) sources without electromagnetic counterparts - offer a powerful route to precision measurements of the Hubble constant, $H_0$, with next-generation (XG) detectors. The statistical promise of this method, however, places stringent demands on waveform accuracy and detector calibration, as even small systematic errors can dominate over statistical uncertainties at high signal-to-noise ratios. We investigate the impact of waveform-modeling systematics on golden dark siren cosmology using a synthetic population of binary black holes consistent with current GW observations and analyzed in the XG-detector era. By comparing state-of-the-art waveform models against numerical-relativity-based reference signals, we quantify modeling inaccuracies from both modeling and data-analysis perspectives and assess how they propagate into biases in luminosity distance, host-galaxy association, and single-event $H_0$ inference. We find that while current waveform models often allow recovery of statistically consistent $H_0$ posteriors, small waveform-induced biases can significantly affect three-dimensional localization and host galaxy ranking, occasionally leading to incorrect redshift assignments. We further derive order-of-magnitude requirements on detector calibration accuracy needed to ensure that calibration systematics remain subdominant for golden dark sirens observed with XG networks. To realize sub-percent $H_0$ measurements with golden dark sirens will require waveform and calibration accuracies that scale as $\mathcal{O}(ρ^{-2})$ with signal-to-noise ratio, motivating sustained advances in waveform modeling, numerical relativity, and detector calibration for the XG era.

We study the absorption time and spreading rate of the discrete-time quantum walk propagating on a line in the presence or absence of an absorber. We analytically establish that in the presence of an absorber, the average absorption time of the quantum walker is finite, contrary to the behavior of a classical random walker, indicating an infinite resource reduction on moving over to a quantum version of a walker. Furthermore, numerical simulations indicate a reversal of this behavior due to the insertion of disorder in the walker's step lengths. Additionally, we demonstrate that in the presence of an absorber, there is a speed-up in the spreading rate, and that a disordered quantum walk that is sub-ballistic regains the ballistic spreading of a clean quantum walk.

We develop a comprehensive quantum estimation framework to quantify how precisely gravitationally induced decoherence can be inferred in optomechanical systems, using single-mode Gaussian probe states. Our approach combines a microscopic description of the gravitational diffusion mechanism with quantum Fisher information to determine the ultimate sensitivity achievable in principle. We show that gravitational diffusion leaves distinct, measurable signatures in the mechanical state, both during transient evolution and in the stationary regime. Finally, we identify how probe state preparation shapes the attainable precision, thereby establishing fundamental limits for detecting and estimating gravity-driven decoherence.

Portfolio optimization under strict cardinality constraints is a combinatorial challenge that defies classical convex optimization techniques, particularly in the context of "Direct Indexing" and ESG-constrained mandates. In the Noisy Intermediate-Scale Quantum (NISQ) era, the Quantum Approximate Optimization Algorithm (QAOA) offers a promising hybrid approach. However, standard QAOA implementations utilizing transverse field mixers often fail to strictly enforce hard constraints, necessitating soft penalties that distort the energy landscape. This paper presents a comprehensive analysis of a constraint-preserving QAOA formulation against Simulated Annealing (SA) and Hierarchical Risk Parity (HRP). We implement a specific QAOA ansatz utilizing a Dicke state initialization and an XY-mixer Hamiltonian that strictly preserves the Hamming weight of the solution, ensuring only valid portfolios of size K are explored. Furthermore, we introduce a Trotterized parameter initialization schedule inspired by adiabatic quantum computing to mitigate the "Barren Plateau" problem. Backtesting on a basket of 10 US equities over 2025 reveals that our QAOA approach achieves a Sharpe Ratio of 1.81, significantly outperforming Simulated Annealing (1.31) and HRP (0.98). We further analyze the operational implications of the algorithm's high turnover (76.8%), discussing the trade-offs between theoretical optimality and implementation costs in institutional settings.

We generalize our previous theorem for FLRW spacetime within the framework of generic metric gravity theories. In our earlier work, we demonstrated that, in the absence of matter fields, the field equations of any generic gravity theory reduce to the Einstein field equations with an effective perfect fluid source. In the present study, we extend this analysis by incorporating scalar and vector fields into a generic gravity theory and show that the resulting field equations remain equivalent to the Einstein field equations with a perfect fluid distribution. We further verify our findings by explicitly examining recently developed Einstein-scalar and Einstein-Proca field theories.

Multi-Agent Path Finding (MAPF) remains a fundamental challenge in robotics, where classical centralized approaches exhibit exponential growth in joint-state complexity as the number of agents increases. This paper investigates Quadratic Unconstrained Binary Optimization (QUBO) as a structurally scalable alternative for simultaneous multi-robot path planning. This approach is a robotics-oriented QUBO formulation incorporating BFS-based logical pre-processing (achieving over 95% variable reduction), adaptive penalty design for collision and constraint enforcement, and a time-windowed decomposition strategy that enables execution within current hardware limitations. An experimental evaluation in grid environments with up to four robots demonstrated near-optimal solutions in dense scenarios and favorable scaling behavior compared to sequential classical planning. These results establish a practical and reproducible baseline for future quantum and quantum-inspired multi-robot coordinations.

Valley splitting in strained Si/SiGe quantum wells is a central parameter for silicon spin qubits and is commonly described with envelope-function and effective-mass theories. These models provide a computationally efficient continuum description and have been shown to agree well with atomistic approaches when the confinement potential is slowly varying on the lattice scale. In modern Si/SiGe heterostructures with atomically sharp interfaces and engineered Ge concentration profiles, however, the slowly varying potential approximation underlying conventional (local) envelope-function theory is challenged. We formulate an exact multi-valley envelope-function model by combining Burt-Foreman-type envelope-function theory, which does not rely on the assumption of a slowly varying potential, with a valley-sector decomposition of the Brillouin zone. This construction enforces band-limited envelopes, which satisfy a set of coupled integro-differential equations with a non-local potential energy operator. Using degenerate perturbation theory, we derive the intervalley coupling matrix element within this non-local model and prove that it is strictly invariant under global shifts of the confinement potential (choice of reference energy). We then show that the conventional local envelope model generically violates this invariance due to spectral leakage between valley sectors, leading to an unphysical energy-reference dependence of the intervalley coupling. The resulting ambiguity is quantified by numerical simulations of various engineered Si/SiGe heterostructures. Finally, we propose a simple spectrally filtered local approximation that restores the energy-reference invariance exactly and provides a good approximation to the exact non-local theory.

We investigate the shadow cast by a rotating traversable wormhole in the Teo class endowed with a general redshift function, with particular emphasis on the emergence of cuspy structures. The shadow boundary is the common envelope of two critical orbit families: unstable circular orbits outside the throat and orbits at the throat itself. The formation of cusps, marking the transition between smooth and cuspy shadow boundaries, only becomes possible when the redshift parameter $λ$ is allowed to vary. Moreover, we uncover a universal critical value $λ_c$ that signals the onset of the cusp. A phase diagram characterized by the spin and redshift parameters reveals four distinct morphologies: smooth, cuspy, ears touching, and throat drowning. The morphology of the wormhole shadow may provide observational diagnostics for the different compact objects in future high-resolution imaging observations.

The formulation of a consistent measurement theory for relativistic quantum fields has become a problem of growing foundational and practical significance. Standard non-relativistic measurement models fail to incorporate the essential relativistic principles of locality, causality, and Lorentz covariance, and are therefore inadequate for quantum field theoretic settings. While most existing work focuses on scalar fields, realistic particles possess spin, polarization, and internal degrees of freedom that introduce new conceptual and operational challenges. To this end, we employ the Quantum Temporal Probabilities (QTP) framework for relativistic measurements to describe electromagnetic, Dirac, and internally structured scalar fields. Our results include probabilities for the time-of-arrival that take spin/polarization into account, generalized photodetection formulas beyond Glauber's theory, an unambiguous derivation of the particle oscillation formula together with its limitations, and a first-principles analysis of relativistic qudits.

We present TensorCircuit-NG, a next-generation quantum software platform designed to bridge the gap between quantum physics, artificial intelligence, and high-performance computing. Moving beyond the scope of traditional circuit simulators, TensorCircuit-NG establishes a unified, tensor-native programming paradigm where quantum circuits, tensor networks, and neural networks fuse into a single, end-to-end differentiable computational graph. Built upon industry-standard machine learning backends (JAX, TensorFlow, PyTorch), the framework introduces comprehensive capabilities for approximate circuit simulation, analog dynamics, fermion Gaussian states, qudit systems, and scalable noise modeling. To tackle the exponential complexity of deep quantum circuits, TensorCircuit-NG implements advanced distributed computing strategies, including automated data parallelism and model-parallel tensor network slicing. We validate these capabilities on GPU clusters, demonstrating a near-linear speedup in distributed variational quantum algorithms. TensorCircuit-NG enables flagship applications, including end-to-end QML for CIFAR-100 computer vision, efficient pipelines from quantum states to neural networks via classical shadows, and differentiable optimization of tensor network states for many-body physics.

This paper presents a comprehensive cryogenic analog signal processing architecture designed for superconducting qubit control and quantum state readout operating at 4 Kelvin. The proposed system implements a complete bidirectional signal path bridging room-temperature digital controllers with quantum processors at millikelvin stages. The control path incorporates a Phase-Locked Loop (PLL) for stable local oscillator generation, In-phase/Quadrature (I/Q) modulation for precise qubit gate operations, and a cryogenic power amplifier for signal conditioning. The readout path features a Low Noise Amplifier (LNA) with 14 dB gain and 8-Phase Shift Keying (8-PSK) demodulation for quantum state discrimination. All circuit blocks are designed and validated through SPICE simulations employing cryogenic MOSFET models at 180nm that account for carrier freeze-out, threshold voltage elevation, and enhanced mobility at 4 K. Simulation results demonstrate successful end-to-end signal integrity with I/Q phase error below 2°, image rejection ratio exceeding 35~dB, and symbol error rate below $10^{-6}$. This work provides a modular, simulation-validated framework for scalable cryogenic quantum control systems.

This work extends previous results on the inverse scattering problem within the framework of Marchenko theory (fixed-$l$ inversion). In particular, I approximate an $n$-channel $S$-matrix as a function of the first-channel momentum $q$ by a sum of a rational term and a truncated sinc series for each matrix element. Relativistic kinematics are taken into account through the correct momentum-energy relation, and the necessary minor generalization of Marchenko theory is given. For energies where only a subset of scattering channels is open, the analytic structure of the $S$-matrix is analyzed. I demonstrate that the submatrix corresponding to closed channels, particularly near their thresholds, can be reconstructed from the experimentally accessible submatrix of open channels.The convergence of the proposed method is verified by applying it to data generated from a direct solution of the scattering problem for a known potential, and comparing the reconstructed potential with the original one. Finally, the method is applied to the analysis of $S_{31}$ $πN$ scattering data.

We investigate the performance of the charger-mediated quantum battery modeled by a two-qubit system. One of the qubits acts as the battery and the other acts as the charger which is subjected to a reservoir. We derived the time-local master equation in Lindblad form with a time-dependent dephasing rate. The dephasing rate may be negative in the early-stage of the charging process and thus indicate the presence of the memory effect. We find that such early-stage memory effect could increase the maximal ergotropy of the battery compared with the one under Markovian approximation with the corresponding asymptotic dephase rate. The enhancement of the performance is explained by means of the non-Markovian quantum jumps. Moreover, a discrete time scheme of the measurement-enhanced quantum battery is proposed in a quantum circuit with global and random local operations.

We review analyses of open quantum systems. We show how non-Hermiticity arises in an open quantum system with an infinite environment, focusing on the one-body problem. One of the reasons for taking the present approach is that we can solve the problem completely, making it easier to see the structures of problems involving open quantum systems. We show that this results in the discovery of a new complete set, which is one of the main topics of the present article. Another reason for focusing on the one-body problem is that the theory permits the strong coupling between the system and the environment. In the current research landscape, it is valuable to revisit the one-body problem for open quantum systems, which can be solved accurately for arbitrary strengths of the system-environment couplings. A rigorous understanding of the problem structures in the present approach will be helpful when we tackle problems with many-body interactions. First, we consider potential scattering and directly define the resonant state as an eigenstate of the Schrödinger equation under the Siegert outgoing boundary condition. We show that the resonant eigenstate can have a complex energy eigenvalue, even though the Hamiltonian is seemingly Hermitian. Second, we introduce the Feshbach formalism, which eliminates the infinite degrees of freedom of the environment and represents its effect as a complex potential. The resulting effective Hamiltonian is explicitly non-Hermitian. By unifying these two ways of defining resonant states, we obtain a new complete set of bases for the scattering problem that contains all discrete eigenstates, including resonant states. We finally mention the non-Markovian dynamics of open quantum systems. We emphasize the time-reversal symmetry of the dynamics that continuously connects the past and the future. We can capture it using the new complete set that we develop here.

Mimetic gravity has emerged as a compelling extension of General Relativity (GR), originally motivated by the attempt to isolate the conformal degree of freedom of the gravitational field. By reparametrizing the physical metric in terms of an auxiliary metric and a scalar field, the theory naturally gives rise to a longitudinal degree of freedom that mimics the behavior of cold dark matter. This review provides a comprehensive survey of the theoretical landscape of mimetic gravity and its multifaceted applications to cosmology and high-energy physics. We begin by examining the original formulation and addressing the fundamental question of its equivalence to GR, highlighting how a singular disformal transformation introduces new physical degrees of freedom. We then explore minimal generalizations that lead to unified cosmological models, including mimetic matter scenarios and extensions into $f(R, φ)$ gravity, which allow for the reconstruction of any desired expansion history. Significant attention is given to the ``limiting curvature'' hypothesis through $f(\Box φ)$ modifications, providing a classical mechanism for resolving cosmological and black hole singularities. We critically assess the challenges facing the theory, specifically the gradient and ghost instabilities identified in cosmological perturbations, and discuss modern resolutions such as ghost-free mimetic massive gravity and covariant formulations of Hořava gravity. Finally, we discuss the role of the mimetic field in the early universe, specifically in the context of asymptotically free gravity and the resolution of the self-reproduction problem in inflation.

We develop a fully discrete and non-perturbative formulation of two-dimensional Jackiw-Teitelboim (JT) gravity within the BF framework. Using group-valued holonomies and Lie-algebra--valued dilatons, the bulk theory is shown to be purely topological, with all physical information encoded at the boundary. We analyze admissible discrete boundary conditions and derive the corresponding asymptotic symmetry algebras directly at the lattice level, including an affine Kac-Moody symmetry and its Brown-Henneaux reduction to a Virasoro algebra, together with the associated Virasoro-dilaton structure. A precise operator product expansion (OPE) dictionary is established by taking the controlled continuum limit of the discrete Poisson brackets. Beyond asymptotic symmetries, we provide an effective boundary description and a representation-theoretic quantization organized by monodromy sectors. Within this discrete framework, black hole entropy follows from gauge-invariant holonomy data and is expressed in terms of the dilaton Casimir, reproducing the Bekenstein--Hawking result without invoking a fundamental Schwarzian action.

Spin squeezing generated via inter-atom entanglement in multilevel atomic ensembles provides a powerful resource for quantum-enhanced metrology. Existing schemes that harness internal atomic degrees of freedom to boost squeezing typically encode the collective squeezing in complex superpositions of magnetic sublevels, which complicates state control and limits practical applications. Here, we propose a coherent control scheme that simultaneously enhances collective spin squeezing and maps the resulting atom-atom entanglement onto two well-defined magnetic sublevels suitable for subsequent metrology experiments. Our protocol sandwiches a quantum non-demolition measurement between two internal one-axis-twisting interactions arranged in an echo sequence. We show that this approach can optimally leverage internal states to boost the inter-atom entanglement and, at the same time, encode it in two magnetic sublevels, which is readily convertible into metrologically useful spin squeezing. Our results offer a straightforward and efficient strategy for generating highly entangled yet readily accessible quantum states in multilevel atomic systems.

Emitter ensembles constitute a fundamental component in quantum optical technologies, yet efficient and accurate simulation of large ensembles remains challenging. Here, we formulate a truncated Wigner approximation (TWA) for permutation-invariant emitter ensembles subject to local dissipation by sampling stochastic trajectories in an extended phase space encompassing the Bloch sphere. Benchmarks show that the TWA accurately captures dynamics, including nonclassical signatures, with the approximation improving with ensemble size. We demonstrate large-scale simulations of hundreds of interacting ensembles within the TWA to reveal emergent spatial coherence and selective directionality of cooperative emission in a pumped 1D chain, highlighting a practical path to studying extended light-matter systems. Our results expand the scope of scalable simulations of quantum emitter ensembles, establishing a bridge between microscopic models and emergent behavior.

We provide an explicit geometric classification of single-qubit projective measurements on one-dimensional linear cluster states within a topological framework. By establishing an explicit geometrical correspondence between local measurements and topological surgery operations on an associated link model i.e. a measurement surgery correspondence, we represent the cluster state as a linear Hopf chain. Within this model, measurements in the computational ($Z$) basis act as topological severance in case of bulk measurements while boundary pruning happens for end measurements of qubits. In contrast, transverse ($X$) basis measurements remove the measured qubit while splicing its neighbours, preserving connectivity through real valued correlations. We show that lateral ($Y$) basis measurements also preserve connectivity but generate intrinsically complex phase factors that are not captured by unframed link models, rendering X and Y measurements topologically indistinguishable at the level of connectivity alone. To resolve this ambiguity, we introduce a framed ribbon representation in which quantum phases are encoded as geometric twists, with chiral $\pm 90^\circ} twists corresponding to the phases $\pm i$. This framing yields a phase-sensitive and outcome resolved topological description of all single qubit measurements on linear cluster states. Our approach provides a unified geometric interpretation of measurement-induced entanglement transformations in measurement-based quantum computation, revealing that quantum phases correspond directly to framed topological invariants. The work is restricted to one-dimensional linear cluster states and single-qubit measurements in the Pauli bases.

Rydberg atomic receivers hold extremely high sensitivity to electric fields, yet their effective 3-dB baseband bandwidth under conventional electromagnetically induced transparency (EIT) is typically constrained to tens to a few hundreds of kilohertz, which hinders wideband wireless applications. To relax this bottleneck, we investigate a six-wave mixing (SWM)-based Rydberg atomic receiver as a wideband radio frequency (RF)-to-optical quantum transducer. Specifically, we develop an explicit baseband input-output model spanning from the probe input to the output light field. Based upon this model, a closed-form 3-dB bandwidth expression is derived to expose its dependence on key optical and RF parameters. We further quantify the linear dynamic range by employing the 1-dB compression point (P1dB) and the input-referred third-order intercept point (IIP3), unveiling a communication-compatible characterization of the bandwidth-linearity trade-off. Finally, our numerical results demonstrate that, given identical optical driving conditions, the SWM configuration increases the 3-dB baseband bandwidth by more than an order of magnitude compared to the EIT-based counterpart, while retaining comparable electric-field sensitivity and revealing a broad, tunable linear operating region.

We present an analysis of the time domain measurement of temporally resolvable quantum states using balanced homodyne detection. Our approach outlines a formalism of detecting quantum states in arbitrary temporal modes via projection of the temporal mode onto a natural detector basis. We then present an algorithm for simulating the resultant photocurrent of continuous homodyne detection in the presence of a temporally resolved mode, and use this algorithm to explore the effects of realistic measurement errors on marginal reconstruction and quantum state tomography. A complete implementation of the method is provided through open source code on a GitHub repository.

This review paper comprehensively examines the influence of spatial torsion on quantum fluctuations from the perspectives of Metric-Affine Gravity (MAG) and the Stochastic Variational Method (SVM). We first outline the fundamental framework of MAG, a generalized theory that includes both torsion and non-metricity, and discuss the geometrical significance of torsion within this context. Subsequently, we summarize SVM, a powerful technique that facilitates quantization while effectively incorporating geometrical effects. By integrating these frameworks, we evaluate how the geometrical structures originating from torsion affect quantum fluctuations, demonstrating that they induce non-linearity in quantum mechanics. Notably, torsion, traditionally believed to influence only spin degrees of freedom, can also affect spinless degrees of freedom via quantum fluctuations. Furthermore, extending beyond the results of previous work [Koide and van de Venn, Phys. Rev. A112, 052217 (2025)], we investigate the competitive interplay between the Levi-Civita curvature and torsion within the non-linearity of the Schrödinger equation. Finally, we discuss the structural parallelism between SVM and information geometry, highlighting that the splitting of time derivatives in stochastic processes corresponds to the dual connections in statistical manifolds. These insights pave the way for future extensions to gravity theories involving non-metricity and are expected to deepen our understanding of unresolved cosmological problems.

We study scalar perturbations of the rotating Teo wormhole embedded in asymptotically Anti-de Sitter (AdS) spacetime and demonstrate that the radial Klein Gordon equation exhibits an emergent conformal structure. The smooth traversable throat induces a logarithmic tortoise coordinate that allows the radial equation to be recast as the quadratic Casimir eigenvalue equation, paralleling the hidden conformal symmetry of the rotating Kerr black hole but arising here in a horizonless geometry. The AdS-Teo spacetime possesses two disconnected timelike AdS conformal boundaries that remain causally connected through the wormhole throat, in contrast to the two-sided eternal AdS black hole where horizons play a central role. Using the emergent conformal symmetry, we construct the near-throat generators, derive the effective potential, and obtain a discrete quasinormal-mode spectrum determined by regularity at the throat and standard AdS boundary conditions at infinity. The AdS embedding further enables a minimal holographic interpretation. As an explicit illustration, we compute an equal-time two-point function in the large-Delta (geodesic) limit from a regulated spacelike geodesic that traverses the wormhole, showing how the bulk geometry couples the two asymptotic boundaries. Together, these results provide a unified description of hidden conformal structure, spectral properties, and boundary correlators in a rotating, horizonless asymptotically AdS wormhole.

We extend a perturbative Dyson-type treatment and discrete-symmetry constraints from the Schrödinger and von Neumann equations to a dephasing Lindblad framework. This work develops further the odd-symmetric formulation -- based on stochastic realism and dual temporal boundary conditions -- from general dynamical considerations to specific tools of quantum mechanics. Applying the resulting scaling relations to published single- and double-diffractive data in $pp$ and $p\bar{p}$ collisions (ISR, UA4, UA5, CDF, D0, ALICE, and E710), we show that single-diffraction cross sections are well described by a three-parameter fit with a relative RMS deviation of $\sim 4\%$, substantially improving upon conventional approximations that neglect decoherence. The extracted decoherence factor is consistently $φ\approx 0.89$, in agreement across SD, DD, and E710-based (direct) estimates, and is naturally interpreted as $φ=1$ for CP-invariant dephasing but $φ<1$ for CPT-invariant dephasing, favouring the latter.

One-dimensional chains of coupled spins are minimal models of strongly correlated quantum matter, and have been proposed as wires for transporting quantum information. In liquids, rapid molecular tumbling averages anisotropic dipolar couplings and leaves effective isotropic scalar $J$-coupling Hamiltonians. At zero- to ultralow-field (ZULF) conditions, differences in frequency between nuclear spins of different types are quenched and the internal Hamiltonians can be closely approximated by an isotropic Heisenberg model. In this work, we present [U-$^{13}$C,$^{15}$N]-butyronitrile as a chemically engineered nuclear spin chain whose full spin-spin coupling network can be determined and validated by combining high-field NMR detection with evolution at ultralow fields. Starting from high-field (16.4 T) NMR spectra of $^1$H, $^{13}$C, and $^{15}$N nuclei, we extract all relevant $J$-couplings within a 12-spin network (four $^{13}$C, one $^{15}$N, and seven $^1$H). We then employ a mechanical field-cycling apparatus to prepolarize the spins at high field, shuttle them into a magnetically shielded region for evolution at <50 nT, and detect signals after returning to high field. Fourier analysis of the ultralow-field evolution yields indirect $J$-spectra that are conceptually analogous to ZULF NMR spectra but measured by a high-field NMR spectrometer. We observe clear spectral features at $J$, 1.5$J$, and 2$J$, in good agreement with simulations using the extracted coupling matrix. Finally, we demonstrate 2D experiments that correlate high-field chemical shifts and, thus, fully map interactions within the molecular spin chain. Our results establish [U-$^{13}$C,$^{15}$N]-butyronitrile as an extremely well-characterized spin chain model system and provide a quantitative Hamiltonian benchmark for future hyperpolarization and quantum-control studies.

Rydberg blockade gates are the most experimentally mature entangling operations in neutral-atom quantum processors, combining fast gate times with simple control, but their performance degrades at larger interatomic separations and remains sensitive to motional and technical noise. Non-blockade gate schemes, such as dark-state and geometric protocols, offer complementary robustness but typically rely on complex and experimentally demanding control. Here we show that quantum optimal control enables non-blockade gate schemes to be implemented using the experimentally established pulse-shaping techniques developed for blockade-based gates. Focusing on the dark-state gate, we construct non-adiabatic implementations that preserve the intrinsic robustness of adiabatic dark-state protocols while achieving gate times comparable to time-optimal blockade gates using only smooth, experimentally feasible pulses. The resulting gates exhibit enhanced resilience to motional coupling, laser noise, and interaction inhomogeneity, particularly near and beyond the blockade radius. This work establishes a practical route to fast, robust two-qubit gates without increased experimental complexity.

These notes introduce quantum computation and quantum error correction, emphasising the importance of stabilisers and the mathematical foundations in basic Lie theory. We begin by using the double cover map $\mathrm{SU}_2 \rightarrow \mathrm{SO}_3(\mathbb{R})$ to illustrate the distinction between states and measurements for a single qubit. We then discuss entanglement and CNOT gates, the Deutsch--Jozsa Problem, and finally quantum error correction, using the Steane $[[7,1,3]]$-code as the main example. The necessary background physics of unitary evolution and Born rule measurements is developed as needed. The circuit model is used throughout.

In this paper we use the Newtonian gravitational potential corrected by non-liner effects to obtain new bounds on graviton mass using non-linear density wave theory (NLDW). This potential differs from the gravitational potential obtained in other modified gravity theories (e.g. the weak field limit of Yukawa gravity, Modified Newtonian Dynamics, non-local theories, $Λ$ cold dark matter..). Using this model, we are able to define wavelength of the non-linear wave as an analytical solution of integrable non-linear differential equation (namely, non-linear Schrodinger equation). Assuming that the wavelength of the non-linear wave represents the graviton Compton wavelength, we have found the corresponding upper bound of graviton mass. We compare obtained result with first assessments of LIGO $\&$ Virgo collaboration and we find they are in a good agreement. Present model used to determine the upper limit of graviton mass is completely independent from other methods published until now. We have compared our result with results obtained using several chosen published methods.

Near-field force measurements at submicron separations can probe Casimir effects and hypothetical short-range interactions, but require cryogenic operation and stable, \textit{in situ} control of separation-dependent backgrounds. We propose a self-calibrating quantum force-gradient sensor in which a ferromagnetic microsphere is Meissner-levitated above a type-I superconducting plane, while a bias magnetic field reproducibly tunes the equilibrium gap for in situ separation scans without mechanical approach. The force gradient is encoded as a resonance-frequency shift tracked by a phase-locked loop, and the motion is read out with a SQUID-coupled, flux-tunable microwave resonator that provides adjustable measurement strength without optical heating. Using the input--output formalism, we derive the conditions for reaching the standard quantum limit (SQL) and identify a counterintuitive scaling law: because displacement-to-flux transduction increases with microsphere size, larger microspheres require fewer photons to reach the SQL, enabling a pathway to macroscopic quantum metrology. We quantify the trade-off between suppression of electrostatic patch potentials (via Au coating) and eddy-current dissipation, project force sensitivities of $\sim 10^{-19}\,\rm{N\,Hz^{-1/2}}$ at millikelvin temperatures, and outline protocols to extract Casimir pressure and constrain Yukawa-type deviations from Newtonian gravity over $0.1$--$10\,μ\mathrm{m}$.

A long-standing goal of nuclear theory is to explain how the structure and dynamics of atomic nuclei and neutron-star matter emerge from the underlying interactions among protons and neutrons. Achieving this goal requires solving the nuclear quantum many-body problem with high accuracy across a wide range of length scales and density regimes. In this review, we discuss how artificial neural network representations of the nuclear many-body wave function have significantly extended the capabilities of continuum quantum Monte Carlo methods. In particular, neural network quantum states enable calculations of larger systems than were previously accessible and provide a flexible framework for capturing phenomena that challenge conventional approaches, including the emergence of nuclear clusters and superfluid phases in dense matter. We highlight recent applications to finite nuclei, infinite nuclear and neutron matter, and dynamical processes relevant to lepton-nucleus and nucleus-nucleus scattering. We also discuss conceptual and methodological connections with condensed matter physics, emphasizing developments in neural network quantum states that bridge strongly correlated systems across disciplines. Together, these developments demonstrate how neural-network methods open new avenues toward unified and accurate descriptions of nuclear structure, matter, and reactions.

Aharonov-Bohm (AB) caging is a complete localization phenomenon in two-dimensional lattices due to destructive interference induced by the background gauge fields. However, current investigations of AB caging are mostly restricted to the Abelian gauge field case, and the observation of AB caging under non-Abelian gauge fields in a quantum system still remains elusive. Here, we report experimental realization of tunable synthetic non-Abelian SU(2) gauge fields in a rhombic lattice, engineered within the synthetic dimensions of a vibrating trapped ion with multiple levels. We realize AB caging under both Abelian and non-Abelian gauge fields and systematically investigate the distinctive transport properties of the non-Abelian case. In particular, we observe typical emergent quantum dynamics unique to non-Abelian AB caging, including initial-state-dependent dynamics, second-order effects, and asymmetric caging behavior. These observations demonstrate the trapped ion system as a powerful platform for simulating emergent phenomena in high-dimensional quantum systems with exotic synthetic gauge fields.

This study offers a comprehensive reconstruction of $f(T)$ gravity model with three distinct non-linear as well as novel forms employing a hybrid scale factor to depict the expansion history of the universe starting from early decelerated epoch to late-time accelerated evolution. Model parameters are rigorously constrained using the Monte Carlo Markov Chain (MCMC) analysis with the help of Bayesian statistics and incorporating late-time observations from BAO and Patheon+SH0ES. The investigation of dynamical parameters such as the equation of state parameter and cosmological parameters indicates alignment with an accelerated expansion phase in both the present and late time epochs. Validation is conducted by assessing the energy conditions, verifying the feasibility of the model forms with particular emphasis on the violation of the strong energy condition that indicates dark energy dominance in modified gravity scenarios. This investigation has been instrumental in determining models that remain consistent with cosmological observations and theoretical requirements. The reconstructed forms of the model effectively mimic $Λ$CDM at late times, providing significant insights into possible extensions of general relativity and bolstering $f(T)$ gravity theory as a robust explanation for cosmic acceleration.

Synthetic frequency dimension offers a powerful approach to simulate lattice models and control photon dynamics. However, extending this concept into the quantum regime, particularly at the single-photon level, has remained challenging in photonic platforms. Here, we demonstrate quantum-state initialization and detection of single-photon evolutions within a synthetic frequency lattice by integrating a superconducting qubit with a 16-meter aluminum coaxial cable. A tunable superconducting quantum interference device (SQUID)-based modulator is employed to synthesize lattice couplings and artificial gauge fields. We observe single-photon quantum random walks and Bloch oscillations, as well as nonadiabatic, unidirectional frequency conversion under rapid temporal modulation of the lattice Hamiltonian, together with band-structure measurements. The lattice connectivity can be readily reconfigured to construct higher-dimensional lattices using multiple drive tones. Our results establish superconducting quantum circuits as a versatile platform for programmable Hamiltonians and extensible synthetic lattices with flexible single-photon control.

Symmetry is often treated in philosophy of physics as an interpretive problem. A particularly lively dispute concerns local symmetries: do they indicate surplus structure that ought to be expunged, or are they merely a harmless redundancy? One influential response favours the second option for certain theories -- those dubbed internally sophisticated. And indeed, in much of physics practice, local symmetries are left implicit: one simply works "up to isomorphism'' without pausing over invariance. But not always. In some settings, local symmetry and invariance become pressing practical concerns for physicists. Yet philosophical discussions of sophistication have paid little sustained attention to when, and why, this happens. Surveying textbook general relativity (GR) and gauge theory, I identify the settings in which diffeomorphism invariance or gauge invariance must be handled explicitly. (Here a setting is a choice of representational framework or background assumptions within which one formulates and uses the theory -- for instance, linearisation, an initial-value formulation, or a Hamiltonian $3+1$ formalism.) I propose an operational criterion -- background-relative sophistication (BRS) -- and argue that it accounts well for the pattern: it marks just where symmetry can stay implicit and where it must be made explicit. Quantum and subsystem settings raise a further difficulty: there, certain tasks (superposition and gluing) force symmetry into view even for theories that are BRS.

We investigate static and dynamical spherically symmetric black hole solutions within the Gravity from Entropy (GfE) framework. We derive and solve the modified vacuum field equations for a static, spherically symmetric spacetime, revealing that the classical Schwarzschild geometry receives perturbative corrections scaling as $r^{-4}$. We establish that the GfE framework is consistent with current strong-field astrophysical observations. Higher-order geometric stresses inherent to the GfE vacuum drive a consistent mass-evolution profile. In the limit of large black hole mass, the theory predicts a constant background evaporation rate $ -β/24$, suggesting an inherent "entropic leakage" of the vacuum. At intermediate scales, the framework replicates the standard Hawking radiation mass-loss law as $\dot{M} \propto M^{-2}$ through a purely classical response of the modified background.

This paper introduces a group-theoretic framework to analyze the algebraic structure of the Grover walk on a complete graph with self-loops. We construct a group generated by the Grover matrix and a diagonal matrix whose entries are powers of a complex root of unity. We then characterize the resulting quotient group, which is defined using a subgroup formed by commutators involving these matrices. We show that this quotient group is isomorphic to a finite cyclic group whose structure depends on the parity of the number of vertices. This group-theoretic characterization reveals underlying symmetries in the time evolution of the Grover walk and provides an algebraic framework for understanding its periodic behavior.

Bridging quantum optics and strong-field physics provides a pathway to explore how quantum light shapes extreme nonlinear light-matter interactions. However, direct characterization of non-classical light at damage-threshold intensities remains an open question. Here, we theoretically investigate the impact of photon-number fluctuations of squeezed light on strong-field photoelectron holography using a quantum-optical strong-field approximation. We identify a mechanism, ponderomotive dephasing, whereby the inherent quantum fluctuations of the driving field dictate the stability of the electron's semiclassical action. While amplitude-squeezed light stabilizes the action to enhance holographic contrast, phase-squeezed light amplifies photon-number noise, causing a rapid collapse of fringe visibility. This quantum-optical sensitivity follows a steep quartic wavelength scaling, rendering mid-infrared drivers uniquely sensitive to the field's underlying quantum nature. Crucially, we show that the collapse of holographic contrast is not a loss of information but a metrological gain. By evaluating the Classical Fisher Information, we identify a "dark-port" mechanism in the tunneling tail that enables the estimation of field quadrature noise beyond the Standard Quantum Limit. This fundamental trade-off between structural imaging fidelity and statistical sensitivity establishes the framework for Attosecond Quantum Tomography: an in-situ, reference-free protocol to reconstruct the Wigner distribution of intense quantum light. Our results identify strong-field ionization as a nonlinear quantum transducer, bridging attosecond electron dynamics with quantum information science.

We show that the claim in Ref. [PRL 131, 200202 (2023)], that the quantum time evolution always can be written as a product of a holonomy operator and a dynamic operator, is false, as it is based on a circular use of the time evolution operator.

We present for the first time a general theory of error correction for permutation invariant (PI) codes. Using representation theory of the symmetric group we construct efficient algorithms that can correct any correctible error on any PI code. These algorithms involve measurements of total angular momentum, quantum Schur transforms or logical state teleportations, and geometric phase gates. For erasure errors, or more generally deletion errors, on certain PI codes, we give a simpler quantum error correction algorithm.

We discuss a scheme to generate large Fock states. The scheme involves repeatedly applying an experimentally feasible unitary transformation to convert a semiclassical state into a Fock state. The transformation combines Kerr interaction, which is a non-Gaussian operation, and pulsed coherent drives. We identify suitable parameter values (Kerr strength, pulse timings, displacement amplitude) for the physical processes to implement the transformation and generate large Fock states with near-unity fidelity. The feasibility of implementing the scheme in circuit QED architectures is discussed. The method is also suitable for generating Fock states of cavity fields.

The Hellings-Downs (HD) correlation serves as the fundamental benchmark for detecting the gravitational-wave background (GWB) in pulsar timing arrays (PTAs) within General Relativity (GR). However, this canonical signature relies on the idealization of a continuum of sources without interference. In realistic astrophysical scenarios dominated by supermassive black hole binaries (SMBHBs), interference between discrete sources induces intrinsic deviations in the spatial correlation, which may mimic or obscure signatures of modified gravity. In this work, we derive the closed-form spatial correlation functions for a GWB with arbitrary polarization and generic GW dispersion relations, in the presence of source interference. Through a rigorous harmonic analysis, we demonstrate that source interference modifies the correlation shape but strictly preserves the lowest non-vanishing multipole moment characteristic of each polarization, specifically the quadrupole for tensor, dipole for vector, and monopole for scalar modes. The truncation at higher-order multipoles is governed by the interplay between pulsar distances and dispersion effects. Furthermore, we quantify the statistical degeneracy between interference-induced variation and modified gravity signatures. We conclude that access to only a single realization of the Universe imposes a fundamental theoretical limit on distinguishing modified gravity from GR using spatial correlations alone.

We study spherically symmetric, self-similar wormhole solutions supported by colliding streams of negative-energy null dust, and their dynamical formation. Under the assumption of self-similarity, the Einstein equations reduce to a system of ordinary differential equations, which we solve numerically under boundary conditions enforcing the existence of a minimal areal radius (the throat) on constant-time hypersurfaces. For a sufficiently large throat radius, the resulting geometries remain regular at both spatial and future null infinity, while a singularity is retained in the past direction. We then construct a dynamical formation scenario by patching together three regions: a Schwarzschild black hole, negative-energy Vaidya spacetimes, and the self-similar wormhole geometry. These regions are joined across null shells using the Barrabes--Israel formalism, which provides explicit relations among the throat radius, the black hole's mass and the energy injection by the shell, demonstrating that an initial black hole can evolve into a wormhole. Our analysis generalizes the formation model for static wormhole solutions proposed by Hayward and Koyama in 2004 to non-static wormhole solutions, offering a novel perspective on the formation of regular traversable wormholes.

In this paper, we develop a theoretical framework that extends single-reference (SR) coupled-cluster (CC) theory beyond its conventional role of describing a single electronic state-typically the lowest-energy state within the symmetry sector defined by the reference determinant. Rather than viewing the SR-CC cluster operator solely as a device for reproducing one target state, we consider more general constructions in which different components of the cluster operator play distinct roles, ranging from encoding states of different symmetry than the reference to enabling SR-CC Ansatz to describe multiple states simultaneously. These developments lead to a new class of SR-CC downfolding formalisms in which the resulting active-space effective Hamiltonians are capable of concurrently representing multiple correlated states nonorthogonal to the reference function. We establish three theorems that formalize this extension and demonstrate that standard CC downfolding emerges as a special case of the proposed framework. Finally, we introduce a Hermitian variant based on a unitary CC representation, which enables realistic simulations of ground and excited states while reducing the quantum resources required.

We present a digitized implementation of rapid adiabatic passage based on a train of weak, frequency-varying ultrafast pulses. Analytic conditions on the subpulse Rabi frequencies and detunings are derived to reproduce the continuous-time population dynamics of a conventional long-pulse excitation. We find that the reproduced dynamics achieves high fidelity even for pulse trains with a small number of subpulses, provided that each subpulse remains within the perturbative regime. The subpulses act as discrete samples of the underlying continuous evolution; consequently, more complex population dynamics, characterized by multiple oscillations prior to the onset of adiabaticity, require a larger number of subpulses for accurate reproduction. In addition, we demonstrate how the sidebands of a frequency comb can be exploited for resonant excitation at large carrier detuning and for the precise preparation of superposition states.

We propose a flux-pumped superconducting parametric amplifier based on symmetrically threaded superconducting quantum interference devices (SQUIDs) that achieves a Kerr-free operating point under suitable drive conditions. Eliminating the Kerr nonlinearity is advantageous for quantum-limited amplification, as it mitigates unwanted distortions in squeezing and prevents degradation of both gain and quantum efficiency in the high-gain strong drive regime. By replacing the central junction in the symmetrically threaded SQUIDs (STS) configuration with a linear inductor, we find that the Kerr-nonlinearity can be eliminated and the effective Hamiltonian reduces to that of a degenerate parametric amplifier (DPA), up to higher-order corrections in the zero-point fluctuations of the superconducting phase operator. We show that the deviations from ideal DPA behavior introduced by these higher-order terms are significantly weaker than those associated with a Kerr nonlinearity. Consequently, the STS design can be driven strongly while maintaining near-quantum-limited performance at the Kerr-free point. Our analysis predicts phase-preserving gain and efficiency approaching the quantum limit, with robust operation demonstrated up to 25 dB of gain.

Entanglement is a central resource in quantum information science; therefore, it is important to design local discrimination protocols that minimize resource consumption. In this paper, we propose three entanglement-allocation schemes for the local discrimination of particular unextendible product bases (UPB) exhibiting strong quantum nonlocality in a \(3 \otimes 3 \otimes 3\) system. By exploiting the structural features of these UPB and the operational advantages of maximally entangled states, we further extend our protocols to strongly nonlocal UPB in \(d \otimes d \otimes d\) systems. In particular, we show that these UPB can be perfectly distinguished with only two maximally entangled states. Moreover, a resource-cost analysis indicates that our protocols, which avoid quantum teleportation whenever possible, can reduce the entanglement consumption. These results not only facilitate resource-efficient quantum information processing, but also provide further insight into the operational role of maximally entangled states.

Integer Linear Programs (ILPs) are a flexible and ubiquitous model for discrete optimization problems. Solving ILPs is \textsf{NP-Hard} yet of great practical importance. Super-quadratic quantum speedups for ILPs have been difficult to obtain because classical algorithms for many-constraint ILPs are global and exhaustive, whereas quantum frameworks that offer super-quadratic speedup exploit local structure of the objective and feasible set. We address this via quantum algorithms for Gomory's group relaxation. The group relaxation of an ILP is obtained by dropping nonnegativity on variables that are positive in the optimal solution of the linear programming (LP) relaxation, while retaining integrality of the decision variables. We present a competitive feasibility-preserving classical local-search algorithm for the group relaxation, and a corresponding quantum algorithm that, under reasonable technical conditions, achieves a super-quadratic speedup. When the group relaxation satisfies a nondegeneracy condition analogous to, but stronger than, LP non-degeneracy, our approach yields the optimal solution to the original ILP. Otherwise, the group relaxation tightens bounds on the optimal objective value of the ILP, and can improve downstream branch-and-cut by reducing the integrality gap; we numerically observe this on several practically relevant ILPs. To achieve these results, we derive efficiently constructible constraint-preserving mixers for the group relaxation with favorable spectral properties, which are of independent interest.

The Discrete Fourier Transform (DFT) is central to the analysis of uniformly sampled signals, yet many practical applications involve non-uniform sampling, requiring the Non-Uniform Discrete Fourier Transform (NUDFT). While quantum algorithms for the standard DFT are well established, a corresponding framework for the non-uniform case remains underdeveloped. This work introduces a quantum algorithm for the Non-Uniform Quantum Fourier Transform (NUQFT) based on a low-rank factorization of the NUDFT matrix. The factorization is translated into an explicit quantum construction using block encodings, Quantum Signal Processing, and the Linear Combination of Unitaries framework, yielding an $ε$-accurate block encoding of the NUDFT matrix with controlled approximation error from both classical truncation and quantum implementation. Under standard oracle access assumptions for non-uniform sampling points, we derive explicit, non-asymptotic gate-level resource estimates. The resulting complexity scales polylogarithmically with target precision, quadratically with the number of qubits through the quantum Fourier transform, and logarithmically with a geometry-dependent conditioning parameter induced by the non-uniform grid. This establishes a concrete and resource-efficient quantum analogue of the NUDFT and provides a foundation for quantum algorithms on irregularly sampled data.

We report a hidden density-wave instability in the trimer-based ruthenate Ba4Ru3O10, previously regarded as a pure antiferromagnet with a phase transition at TA=100 K. This transition is manifested in lattice parameters, transport, thermodynamics, and magnetic susceptibility, yet remains remarkably insensitive to magnetic fields up to at least 14 T, indicating an electronically driven reconstruction. At much lower temperatures T*= 20 K, charge transport becomes strongly nonlinear, exhibiting distinct depinning thresholds, negative differential resistance, pronounced current- and frequency-dependence, and slow collective dynamics in the Hertz range. While each feature is characteristic of density-wave transport, their simultaneous occurrence in an antiferromagnetic oxide is unprecedented. All nonlinear signatures vanish upon only 3% Ir substitution, which preserves the crystal structure and insulating state, ruling out Joule heating or extrinsic artifacts. The wide separation between the electronic reconstruction at TA and the emergence of nonlinear dynamics at T* identifies Ba4Ru3O10 as a rare correlated system hosting a strongly pinned collective electronic state intertwined with antiferromagnetism.

Effective Field Theories with higher derivatives often yield equations of motion which define ill-posed problems. We present a method for enhancing control on such theories by coupling them to a field living in one extra dimension. The resulting action principle helps to define a well-posed problem introducing a mechanism to control UV behavior. Physically this is achieved by dissipating the energy in the short-wavelength modes into the extra dimension. We examine the resulting dynamics and compare it to alternative proposals for studying such theories in the non-linear regime.

In a previous work we developed a field theoretical approach to open quantum systems using condensed matter methods. In the Born approximation we derived the Redfield equation on the basis of a multi-oscillator bath, a Dyson equation, a diagrammatic expansion and a quasi-particle approximation. In addition applying a rotating wave approximation we obtained the Lindblad equation describing a proper quantum map. The issue regarding the additional rotating wave approximation was left as an open problem. The present work addresses the open problem and presents new results. We identify a discrepancy in the popular and standard Redfield equation. The discrepancy is associated with the well-known fact that the Redfield equation does not represent a proper quantum map. The discrepancy is related to the diagrammatic expansion and a consistency requirement in the quasi-particle approximation. The explicit resolution of this discrepancy is obtained by imposing energy conservation on the Born level. As a result we obtain formal equivalence between the energy-conserving Redfield equation and the Lindblad equation without invoking the rotating wave approximation. We provide a detailed mapping of the field theoretical approach to the standard microscopic derivation in the theory of open quantum systems.

Quantum error correction offers a promising path to suppress errors in quantum processors, but the resources required to protect logical operations from noise, especially non-Clifford operations, pose a substantial challenge to achieve practical quantum advantage in the early fault-tolerant quantum computing (EFTQC) era. In this work, we develop a systematic scheme to encode exponential maps of the form $\exp(-iθP)$ into stabilizer codes with simple circuit structures and low qubit overhead. We provide encoded circuits with small first-order logical error rate after postselection for the [[n, n-2, 2]] quantum error-detecting codes and the [[5, 1, 3]], [[7, 1, 3]], and [[15, 7, 3]] quantum error-correcting codes. Detailed analysis shows that under the level of physical noise of current devices, our encoding scheme is 4--7 times less noisy than the unencoded operation, while at most 3% of runs need to be discarded.

Identifying stabilizer codes that admit fault-tolerant implementations of the full logical Clifford group would significantly advance fault-tolerant quantum computation. Motivated by this goal, we study several classes of fault-tolerant gadget constructions consisting of Clifford gates acting on the physical qubits, including transversal gadgets, code automorphisms, and fold-transversal gadgets. While stabilizer codes encoding a single logical qubit, most notably the [[7,1,3]] Steane code, are known to admit transversal implementations of the full logical Clifford group, no analogous examples are known for codes encoding multiple logical qubits. In this work, we prove a no-go theorem establishing that no stabilizer code admits a fully transversal implementation of the Clifford group on more than one logical qubit. We further strengthen this result by showing that fold-transversal implementations of the full logical Clifford group are impossible for stabilizer codes encoding more than two logical qubits. More generally, we introduce the notion of k-fold transversal gadgets and prove that implementing the full Clifford group on k logical qubits requires at least k-fold transversal gadgets at the physical level. In addition, we analyze code-automorphism based constructions and demonstrate that they also fail to realize the full Clifford group on multiple logical qubits for any stabilizer code. Together, these results place fundamental constraints on fault-tolerant Clifford gadget design and show that stabilizer codes supporting the full logical Clifford group on multiple logical qubits via these architectures do not exist. Since the Clifford group is a core component of universal gate sets, our findings imply that quantum computing with codes encoding multiple logical qubits within a single code block necessarily entails more complex constructions for fault tolerance.

Cold dark matter may be a fluid (or plasma) residing in a strongly-interacting hidden sector, rather than a population of weakly-coupled particles. Such a scenario admits a holographic description in terms of a cosmological braneworld embedded in the linear dilaton five-dimensional (5D) spacetime. In this framework, dark matter originates from the linear dilaton bulk black hole, whose phase we show to be thermodynamically favored at all temperatures. We present a natural freeze-in mechanism for the production of holographic dark matter, in which the bulk black hole is fed by energy leaking from the brane after inflation. Our model is characterized by two free parameters, one of which, the position of the black hole horizon, is fixed by the observed dark matter abundance. The remaining parameter, the 5D Planck scale $M_5$, is consistent with all current experimental bounds provided that $M_5\gtrsim 3\times 10^5$ TeV.

Recent developments in gravitational path integrals indicate that the nonperturbative physical Hilbert space of a closed universe is one-dimensional within each superselection sector. This raises a basic puzzle: how can a unique quantum-gravity state give rise to semiclassical physics, measurement outcomes, and classical probabilities? In this paper, we develop a framework in which nontrivial and statistically stable predictions emerge despite the one-dimensionality of the fully constrained Hilbert space. The key idea is to extract physical predictions in an enlarged, unconstrained Hilbert space by conditioning on observational data. We show that partial observability -- reflecting the limited access of observers to the degrees of freedom of the universe -- suppresses ensemble fluctuations associated with microscopic structure in the gravitational path integral, thereby restoring semiclassical predictability with exponential accuracy. We formulate the construction explicitly including contributions from the Hartle--Hawking no-boundary state, define a gauge-invariant Hilbert space for observations via a density operator, and generalize the formalism to conditioning on histories, clarifying the emergence of classical probabilities and an effective arrow of time. Finally, we explore whether this framework can support a realistic cosmology and identify assumptions that the underlying theory of quantum gravity must satisfy.

Simulating strongly correlated systems in two dimensions is notoriously challenging due to rapid entanglement growth and frustration. Here, we introduce the adaptive projected-purified pseudoboson density-matrix renormalization group (A3P-DMRG) tailored to explore the ground states of dilute lattice models. The method compresses cluster Hilbert spaces by retaining only the most probable low-occupation Fock states, identified via probabilistic bounds and refined through a self-consistent mean-field basis optimization. We demonstrate that A3P-DMRG is advantageous in low-filling and weak-coupling regimes for large system sizes where conventional DMRG struggles. This establishes the method as a versatile tool for studying dilute quantum many-body systems relevant to ultra-cold atom quantum simulators, photonic lattices, Moiré materials and quantum chemistry.

Neutrino oscillations probe the quantum gravity interface in unique ways. While gravitational effects on neutrinos are well studied in general relativity and torsion based geometries, the symmetric teleparallel regime where gravity stems solely from non-metricity, with zero curvature and torsion has remained uncharted. In this work, we perform the first analysis of neutrino oscillations in such a spacetime. Using the reduced Kerr metric in coincident gauge for the slowly rotating and weakly gravitating spherical Sun, we derive the Dirac Hamiltonian from the generalized Dirac equation and compute the accumulated phase of neutrino mass eigenstates. There are six free coupling constants in our model. Based on certain observational inputs, we inferred upper bounds on our arbitrary coupling constants. This allowed us to simplify the otherwise cumbersome calculations to some extent. Ultimately, we computed the phase differences that play a crucial role in solar neutrino oscillations and analyzed the contributions arising from our arbitrary coupling constants. Our results establish neutrino oscillations as a novel probe of non-metricity and open a new avenue for testing symmetric teleparallel gravity through astrophysical observations.

Final states comprising two or more top quarks are important search channels at the Large Hadron Collider for scalar particles predicted in models of physics beyond the Standard Model. While the di-top final state profits from a higher signal cross section, it can be subject to intricate interference patterns. Besides the interference with the large QCD background, in case of the presence of more than one high-mass scalar also large signal--signal interference contributions can occur. We show that in such scenarios it is crucial to account for loop-level mixing for obtaining accurate exclusion bounds. We demonstrate how the interference patterns can obscure the interpretation of possible deviations from the Standard Model expectations. We show that the four-top final state, while giving rise to a smaller signal cross section, provides important complementary information due to its much smaller signal--background interference contributions. Thus, the results obtained from the four-top final state can be instrumental for pinpointing the underlying new physics scenario.

Primordial black holes (PBHs), produced in the early Universe, can source a stochastic background of induced gravitational waves (GWs) and provide a non-thermal origin for dark matter (DM). We investigate DM production in a PBH-dominated cosmological framework, including contributions from PBH evaporation, gravitational production, and thermal freeze-in and freeze-out mechanisms, and determine the regions consistent with the observed DM relic abundance. We find that thermal freeze-in can compensate for the underabundance of PBH-sourced DM, while indirect detection remains largely insensitive due to the feeble interaction strength, making future GW observatories such as LISA and the Einstein Telescope (ET) unique probes of this scenario. For freeze-out DM, indirect detection experiments constrain regions with relatively large annihilation cross-sections, whereas GW observations probe complementary regions with heavier DM masses and smaller interaction strengths. Consequently, the same DM parameter space cannot be simultaneously probed by both indirect detection searches and GW missions. These results establish GW observations as a powerful and independent probe of DM production in PBH-dominated cosmologies, opening a new observational window into DM properties and the thermal history of the pre-BBN Universe.

Fuzzy sphere models conjecturally realize 3d CFTs in small systems of spinful fermions, but why they work so well is still not fully understood. Their Hamiltonians are built from electron density operators projected to the lowest Landau Level. We analyze the algebra of the density modes and verify that it satisfies the Jacobi identity. The fuzzy sphere geometry admits two thermodynamic limits: a local planar limit yielding the fuzzy plane, and a commutative limit yielding an ordinary sphere. In the planar limit, high-angular-momentum modes recover the Girvin-MacDonald-Platzman algebra, whereas in the commutative limit the low-angular-momentum modes become semiclassical. We further find an explicit representation of the conformal algebra so(3,2) in the minimal two-electron system and extend it to larger systems via an so(3) equivariant coproduct. Because the coproduct splits one so(3) representation into a tensor product, it is structurally mismatched with the thermodynamic limit of critical fuzzy sphere Hamiltonians.

Gravitational waves (GWs) produced during first-order phase transitions (FOPTs) in the early universe provide a powerful probe of nonstandard cosmological histories. We study GW production from a FOPT ending a kination-dominated epoch in the Kination-Induced Big Bang scenario, in which a period of kination domination terminates through a phase transition that reheats the universe into radiation domination. A rolling scalar field drives the kination epoch. In the specific model we consider, its derivative coupling to a second scalar (tunneling field) dynamically traps the latter in a false vacuum, with the phase transition triggered as the kination field slows due to Hubble friction. We compute the resulting stochastic GW background from bubble nucleation and collisions, presenting analytic estimates and numerical results for the peak amplitude and frequency. In all cases we find an upper bound $Ω_{\rm GW} h^2\lesssim 2\times10^{-7}$ from the bubble percolation condition. In the case where the false vacuum energy dominates at the transition (yet the kination field drives the FOPT), we find $Ω_{\rm GW}h^2\gtrsim 10^{-12}$. We further find that the Hubble scale during the phase transition across a broad set of model parameters is bounded by $\mathscr{O}(10^{-13})M^2/M_{\rm Pl}\lesssim H_* \lesssim \mathscr{O}(0.1)M^2/M_{\rm Pl}$, where $M$ is the mass-scale controlling the strength of the interaction between the kination and tunneling fields. The predicted signal spans frequencies from nHz to MHz, allowing the model to explain the signal reported by Pulsar Timing Array experiments and to be constrained or probed by interferometers such as LISA, Advanced LIGO, Cosmic Explorer, and BBO. Interestingly, a FOPT can occur even if the bare tunneling potential has a single minimum, as metastability is generated dynamically by the coupling between the tunneling and the kination field.

Quantum electrodynamics (QED), the first relativistic quantum field theory, describes light-matter interactions at a fundamental level and is one of the pillars of the Standard Model (SM). Through the extraordinary precision of QED, the SM predicts the energy levels of simple systems such as the hydrogen atom with up to 13 significant digits, making hydrogen spectroscopy an ideal test bed. The consistency of physical constants extracted from different transitions in hydrogen using QED, such as the proton charge radius $r_\mathrm{p}$, constitutes a test of the theory. However, values of $r_\mathrm{p}$ from recent measurements of atomic hydrogen are partly discrepant with each other and with a more precise value from spectroscopy of muonic hydrogen. This prevents a test of QED at the level of experimental uncertainties. Here we present a measurement of the 2S-6P transition in atomic hydrogen with sufficient precision to distinguish between the discrepant values of $r_\mathrm{p}$ and enable rigorous testing of QED and the SM overall. Our result $ν^{}_{\text{2S-6P}}$ = 730,690,248,610.79(48) kHz gives a value of $r_\mathrm{p}$ = 0.8406(15) fm at least 2.5-fold more precise than from other atomic hydrogen determinations and in excellent agreement with the muonic value. The SM prediction of the transition frequency (730,690,248,610.79(23) kHz) is in excellent agreement with our result, testing the SM to 0.7 parts per trillion (ppt) and, specifically, bound-state QED corrections to 0.5 parts per million (ppm), their most precise test so far.

Spectral distortion anisotropies of the cosmic microwave background (CMB) provide a new probe of the early Universe that can be accessed using traditional CMB imaging techniques. It is possible to compute the creation and evolution of anisotropic signals for various scenarios using the frequency hierarchy method recently developed for CosmoTherm. However, the current treatment is not perfect and some approximations had to be made. Here, we carefully construct a modified form for the evolution equations that has the full equilibrium solutions built into the formulation. We improve the formalism to account for i) additional stimulated scattering effects, ii) kinematic corrections to the thermalization terms, iii) corrections to the standard perturbation variables and iv) direct photon sources. These effect could not be captured with the original formulation of the frequency hierarchy method but are indeed important for cleanly separating real distortions from temperature signals. However, we show that previous results are not altered significantly when compared to the improved formulation presented here. As a new worked example, which could indeed not be treated before, we also illustrate how possible changes in the temperature-redshift relation would create spectral distortion anisotropies in the pre-recombination era. The theoretical methods presented here are also an important step towards being able to consistently predict the CMB spectral distortion anisotropies in photon-dark photon and photon-axion conversion scenarios.

A two-parameter class of higher-dimensional, strongly coupled, confining field theories in the presence of magnetic fluxes for two Abelian gauge groups admits a top-down, holographic dual description. The corresponding two-parameter family of regular background solutions of the classical equations of maximal supergravity in seven dimensions descends from maximal supergravity in eleven dimensions. We study the global and local stability properties of these solutions. We identify lines of zero-temperature first-order phase transitions, describing a polygon (a square) in the space of parameters, identified with the two fluxes. In the spectrum of fluctuations of the supergravity equations, interpreted as bound states of the dual, confining field theories, we find no evidence of local instabilities (tachyons). Over a significant portion of parameter space, that extends far away from the proximity to the transition, we identify an approximate dilaton, the mass of which is one order of magnitude smaller than the scale set by confinement. Our findings complement those emerging in other holographic models discussed in the literature, in which either the dilaton mass is only mildly lower than the confinement scale (when approaching a first-order transitions), or parametrically suppressed (when reaching the proximity to a second-order one).

We use the Born--Oppenheimer effective field theory factorization to compute the inclusive production cross sections of the $χ_{c1}(3872)$ and its partner in the bottomonium sector. In the same framework, we compute the production cross sections of the pentaquark states $P_{c\bar{c}}(4312)^+$, $P_{c\bar{c}}(4457)^+$, $P_{c\bar{c}}(4380)^+$ and $P_{c\bar{c}}(4440)^+$ within two possible scenarios for the Born--Oppenheimer potentials. Also for pentaquarks, we extend the results to the bottomonium sector. All our results are genuine predictions that do not involve fits to prompt hadroproduction data.

CP-odd observables are identified in $γ^*\to K^+K^-π^0$ and evaluated in the chiral limit. A nonzero T-odd asymmetry $A_T$ requires anomalous couplings descending from a five-dimensional Chern-Simons construction. Modeling the running of $F_A$ and $F_V$ with the vector meson dominance, we estimate sensitivities to the strange-quark EDM: $\mathcal{O}(10^{-16}),e\cdot\mathrm{cm}$ in $e^+e^-$ data at VEPP-2000 (CMD-3) and $\mathcal{O}(10^{-18}),e\cdot\mathrm{cm}$ using existing $J/ψ$ samples at BESIII. Future experiments in the Super Tau-Charm Facility and Belle II can further improve the reach by an order of magnitude.

We complete the determination of the parapositronium decay into three photons by evaluating amplitudes mediated by the $Z$ boson. We show that, contrary to the expectation that the extra mass scale $m_e$ may bring an enhancement to the overall scaling, the amplitude turns out to start at $1/m_Z^6$ order, similarly to the $W$ boson mediated amplitude. The decay rate with both $W$ and $Z$ boson contributions is found to be $1.1\cdot 10^{-80}$ eV, about 47 orders of magnitude smaller than previously estimated. We also discuss its implication for the $π^0\to3γ$ amplitude.

Searches for stochastic gravitational wave backgrounds from first-order phase transitions offer a powerful probe of hidden sectors, but quantitative predictions in gauge theories are obstructed by the gauge dependence of the finite-temperature effective potential and the associated tunneling action. We study a minimal gauged $U(1)$ dark sector containing a dark Higgs and a dark photon, optionally supplemented by a vectorlike dark fermion, coupled to the Standard Model through Higgs portal or kinetic mixing. Using the Nielsen identity together with a controlled derivative expansion and power counting, we construct a gauge-independent effective action in the high- and low-temperature limits, enabling model-intrinsic nucleation dynamics and robust gravitational wave predictions. We perform dedicated Monte Carlo scans in both limits and map viable microscopic parameters to detector-facing peak frequencies and amplitudes, spanning bands relevant to pulsar timing arrays and planned space-based interferometers. In our scans, supercooled transitions typically produce the strongest signals, whereas parametrically high-temperature transitions are comparatively rare and tend to be weak. We further connect the phase transition phenomenology to viable dark matter candidates within the same minimal field content, providing benchmark targets for dark photon dark matter and dark fermion dark matter, and highlighting their complementarity with gravitational wave observables. Overall, our results provide an end-to-end, gauge-independent pipeline from a minimal hidden sector Lagrangian to gravitational wave spectra and cosmologically viable dark matter benchmarks, yielding the most reliable and concrete predictions to date for a minimal gauged $U(1)$ dark sector.

Light scalar and vector particles admit non-trivial descriptions in terms of anti-symmetric higher-rank tensor fields. Far from mere rewritings, these provide compelling alternative frameworks, leading to immediate phenomenological applications. In this paper, we extend the playground to include the contributions of fermionic triangle loops, and use these results to compare the standard and higher-form realizations for two phenomenological processes: the pair production of two dark particles and the decay of a dark particle in two photons. Though they all do show some dependencies on the chosen realizations for the spin-zero or the spin-one dark states, we find that the two-photon mode is particularly sensitive and could actually be our prime window into the true nature of the dark field.

We introduce a renormalization-group invariant observable, the symmetry strength parameter $κ_{AB}$, for the quantitative characterization of symmetry breaking in QCD. As a first application, we employ $κ_{AB}$ to investigate the relative strength of $SU(2)_L \times SU(2)_R$ chiral symmetry and $U(1)_A$ axial symmetry breaking in $N_f=2+1+1$ lattice QCD using optimal domain-wall fermions at the physical point. Our study covers three lattice spacings and twelve temperatures in the range 164-385~MeV. We examine three independent symmetry-breaking channels in the nonsinglet sector with connected correlators: the $U(1)_A$-sensitive scalar-pseudoscalar channel ($κ_{PS}$), probing the $π$-$δ$ system; the $SU(2)_L \times SU(2)_R$-sensitive vector--axial-vector channel ($κ_{VA}$), probing the $ρ$-$a_1$ system; and an additional $U(1)_A$-sensitive tensor--axial-tensor channel ($κ_{TX}$), probing the $ρ$-$b_1$ system. At finite lattice spacing, we observe a clear hierarchy $κ_{PS} > κ_{TX} \sim κ_{VA}$. A controlled continuum extrapolation reveals that this hierarchy collapses, with all three symmetry-breaking strengths becoming statistically indistinguishable within our precision. This result provides a new, model-independent benchmark from a chirally symmetric lattice action. Our findings indicate that the effective restoration scales for $SU(2)_L \times SU(2)_R$ and $U(1)_A$ in the nonsinglet sector converge closely near the chiral crossover, placing stringent quantitative constraints on the temperature window for chiral and axial symmetry manifestation in connected channels. These results support a two-stage restoration scenario, in which full symmetry restoration -- including the singlet sector -- occurs only at significantly higher temperatures once topological fluctuations are suppressed.

The excavation of the Hyper-Kamiokande cavern, 600 m underground, is complete. Measuring 69 m in diameter and 94 m in height, it is among the world's largest rock caverns. A vertically oriented, dome-capped cylindrical design was chosen to optimize cost and performance. Combined with substantial overburden, the geometry posed major engineering challenges. This paper outlines the underground works, main cavern design, excavation plan, and the evolution of support design and construction methods during excavation, namely the information-based (observational) design and construction approach.

Light-cone distribution amplitudes (LCDAs) are essential to precision phenomenology in heavy baryon decays. In this work, we derive a factorization formula connecting the leading-twist QCD LCDA to the boosted HQET LCDA of the $Λ_Q$ baryon in the peak region. We demonstrate a significant simplification of the matching procedure by applying the method-of-regions to perturbative calculations. With this simplification, we calculate the required one-loop perturbative corrections to the QCD and boosted HQET LCDAs in the $\overline{\rm MS}$ scheme, and thereby obtain the one-loop jet function that serves as the matching kernel in the factorization formula. This result provides a critical step toward lattice QCD calculation of heavy baryon LCDAs in the future.

We present the first rainbow-ladder Dyson-Schwinger equations study of the pion's chiral-odd, twist-3 parton distribution $e_{\rm q}(x)$. By deriving a novel Dyson-Schwinger equation for the pion's quark-quark correlation matrix, we simultaneously extract from it both the unpolarized twist-2 parton distribution function $f_{\rm q}(x)$ and the twist-3 distribution $e_{\rm q}(x)$. Our results show that chiral symmetry strongly suppresses the twist-2 component of $e_{\rm q}(x)$, leaving the genuine twist-3 quark-gluon component dominant and featuring a node. We therefore argue that the genuine twist-3 term can make a substantial contribution to hadronic $e(x)$, producing a nonmonotonic structure that is a clear imprint of intrinsic quark-gluon correlations. We point out that the existence of a hump-like feature is compatible with, although not uniquely indicated by, recent proton extractions and awaits more precise determination.

Anomalous dimensions of leading-twist operators in QCD play an important role in precision predictions for high-energy processes, since they govern the scale evolution of parton distributions. Their analytic structure as a function of spin is particularly important due to the complexity of higher-loop computations. In these proceedings, we discuss the resummation of the certain type of such singularities that share common features with those appearing in the quark flavor-nonsinglet sector of QCD. Our main focus is on the interplay between Gross-Neveu-Yukawa model in $ε$ expansion and Gross-Neveu in $1/N$ expansion. Such resummation allows one to predict the higher-loop singular behavior and reveals connections with the conformal Regge theory and recent studies of detector operators in QCD and various conformal field theories.

We propose a scale invariant radiative neutrino mass model with custodial symmetry. Masses of an inert doublet scalar and right-handed neutrinos are induced by a vacuum expectation value (VEV) of a singlet scalar caused through the Coleman-Weinberg mechanism. It violates spontaneously both the custodial symmetry and the scale invariance. The weak scale can take a suppressed value compared with the singlet scalar VEV because of the custodial symmetry. In this framework we study phenomenological consequences for neutrino mass, dark matter and baryon number asymmetry. Since the required dark matter abundance cannot be explained by a neutral component of the inert doublet scalar, the lightest right-handed neutrino should be dark matter. Mass of the dark matter is predicted to be less than $O(1)$ MeV and baryon number asymmetry could be explained through resonant leptogenesis.

The dense matter equation of state is of great interest due to the recent development of astrophysical observations for neutron stars. A rapid increase in pressure indicates a continuous crossover from a hadron phase to a quark phase without any phase transitions, yet its microscopic mechanism remains elusive. Recently, a peak in the speed of sound and a baryon momentum-shell structure, which are predicted from a quarkyonic matter picture, have been regarded as key features of the hadron-quark crossover. In this work, we explore a field-theoretical framework to describe the hadron-quark crossover, drawing an analogy with the Bose-Einstein condensate to Bardeen-Cooper-Schrieffer (BEC-BCS) crossover established in ultracold atomic experiments. Strikingly, a peak in the speed of sound and the baryon momentum-shell structure can simultaneously be explained by the tripling fluctuation effect arising from a different context of quantum many-body physics. We demonstrate these properties in a simplified model and provide a microscopic derivation of the quarkyonic matter model within our field-theoretical framework.

Given recent advancements in dark matter (DM) search experiments, particularly the latest LUX-ZEPLIN (LZ) direct detection (DD) results, we systematically investigate the light bino-higgsino DM scenario within the natural supersymmetric framework. Requiring the electroweak fine-tuning parameter $Δ_{\text{EW}} < 30$ fixes the higgsino mass parameter in the range $|μ| \in [100, 350]$ GeV, while we extend the bino mass to $M_1 \in [10, 350]$ GeV. Incorporating constraints from Higgs physics, rare $B$ decays, LEP limits, and DD experiments, we find that part of the parameter space remains viable. However, the relic density of neutralino DM necessarily lies below the observed Planck value, contributing at most $\sim 2\%$ of the total DM abundance. Some of the surviving parameter space is already excluded by current 13 TeV LHC searches, while the future 14 TeV HL-LHC with 3000 fb$^{-1}$ luminosity will probe the remaining region.

The data acquisition system of the upgraded LHCb experiment includes the fast reconstruction of all hits in the vertex locator (VELO) pixel detector at the beam-crossing rate of 40 MHz, implemented as on-the-fly clustering embedded in the firmware of the readout board FPGAs. The availability of a high rate of reconstructed clusters in real time enables a new fast approach for measuring luminosity and monitoring the LHCb luminous region, directly at the detector readout level. This methodology has been implemented as an array of real-time cluster counters in the VELO readout FPGAs and has been in operation since the start of the 2024 physics run of LHCb. This paper describes the methodology and its features and performance, both on proton-proton and lead-lead collision data. The method shows a statistical resolution better than the percent level, and a sensitivity to variable running conditions of the same level. This is achieved with an intrinsic time granularity better than 100 ms , undersampled to 3 s for analysis purposes. Nonlinear behaviour is compatible with zero in a luminosity range including the LHCb Run 3 operating point.

This study investigates the excesses observed around 95 GeV in diphoton and $b\bar{b}$ experiments within the framework of the Left-Right Supersymmetric Model (LRSSM). Considering the one-loop and two-loop effective potential corrections to the Higgs masses, the model is able to describe the experimentally observed $μ(h_{95})_{γγ}$ and $μ(h_{95})_{b\bar{b}}$ signal strengths. In addition, we also present the impacts of the LRSSM-specific parameters $\tanβ_{R}$, $v_{R}$ and $v_{S}$ on the theoretical predictions of the signal strengths for the 95 GeV and 125 GeV neutral Higgs both.

In this work, we compute the next-to-leading-order QCD corrections to the Dirac electromagnetic form factors of the $Σ$ hyperons within the hard-collinear factorization framework at leading power. The corresponding short-distance coefficient functions are extracted from the relevant seven-point partonic correlation functions. We find that the one-loop radiative corrections to the leading-twist hard-scattering contributions are numerically significant over a broad range of momentum transfer. Combining the perturbatively calculated hard kernels with nonperturbative $Σ$ distribution amplitudes determined from lattice QCD, we present state-of-the-art theoretical predictions for the $Σ$ hyperon electromagnetic form factors.

The counting of microstates of supersymmetric black holes with anti-de Sitter or flat asymptotics is obtained by computing a supersymmetric index in a weakly coupled string theory or a dual superconformal field theory. These indices are protected observables, whose value can be reliably extrapolated from weak to strong coupling, where the gravitational description applies. In this Thesis, after a broad introductory review, we discuss recent progress in formulating such protected observables directly within the gravitational theory, via the Euclidean path integral. In the semiclassical limit the index reduces to a sum over complex Euclidean saddles. These saddles are supersymmetric but ''non-extremal'', and arise in both anti-de Sitter and flat spaces. In the holographic setting, we investigate four-derivative corrections to the thermodynamics of AdS$_5$ black holes. Using off-shell methods, we construct the corrected action of five-dimensional gauged supergravity. We then evaluate the corrected on-shell action of supersymmetric AdS$_5$ black holes and find exact agreement with a Cardy-like limit of the superconformal index of the dual conformal field theory. By a Legendre transform of the action, we obtain the corrected black hole entropy, and we confirm this result by applying Wald's formula to the corrected near-horizon geometry. We then turn to the gravitational index with asymptotically flat boundary conditions. We uncover a broad family of novel saddles and present a general classification based on their rod structure, which characterizes their topology. These solutions may feature multiple horizons or three-dimensional bubbles with lens space topology, and allowing for conical singularities yields further geometries, involving spindles and branched spheres. For the simpler geometries, the on-shell action is computed using an odd-dimensional version of equivariant localization.

In this paper, we investigate gravitational interactions of massive fields with arbitrary integer and half-integer spin, trying to construct a vertex that contains both standard minimal and non-minimal interaction terms necessary to make the vertex gauge invariant. We propose an ansatz for these non-minimal terms and show that it leads to a unique solution that correctly reproduces our previous results for spins 5/2, 3 and 7/2, including all possible partially massless limits.

We define a 2-form connection in the space of classical solutions of the bosonic open string field theory, using the open string star product and integration. The corresponding higher holonomies and the 3-form curvature are new observables invariant under the infinite-dimensional gauge algebra of open string field theory. The definition is analogous to that of Berry phase in quantum mechanics and is motivated by recent studies on higher Berry phase in condensed matter physics and quantum field theory. We suggest identifying this 2-form connection with the Kalb-Ramond $B$-field of the closed string background at least in favorable situations. Also discussed are sigma models whose target space is the moduli space of conformal boundary conditions of a two-dimensional CFT with the $B$-field given by a cousin of this 2-form connection.

The findings on the spin polarization of $Λ$, $Ξ$, and $Ω$ hyperons and spin alignment of $K^{*0}$, $φ$, and $D^{*+}$ mesons in relativistic heavy-ion collision experiments at the RHIC and LHC facilities propose the emergence of a strong vorticity field produced in these collisions. Contemplating the potential impact of vorticity on the space-time evolution of deconfined QCD matter and its freeze-out properties, we aim to investigate its characteristics within the medium. We introduce a complementary and data-driven approach to quantify the global vorticity field by extracting it directly from the transverse momentum spectra of produced hadrons. Employing the experimental data for $Λ$, $Ξ$, $Ω$, $K^{*0}$, $K^{*\pm}$, $φ$, $ρ$, and $D^{*+}$ at mid-rapidity in Au+Au and Pb+Pb collisions over a wide range of beam energies, $\sqrt{s_{\rm NN}}=7.7$ GeV-5.02 TeV, and centrality classes, we systematically examine spin-vorticity coupling in the medium. Our finding on the magnitude of the extracted vorticity is consistent with values deduced from $Λ$ and $\barΛ$ polarization measurements using statistical thermal models under the non-relativistic limit. Notably, we observe a prominent particle-species dependence of the vorticity, as well as a non-trivial variation with collision centrality and beam energy. These results indicate that vorticity-driven spin phenomena are sensitive to hadron structure and freeze-out dynamics, providing new constraints on the rotational properties of the QCD matter.

We develop a framework of zero dimensional defects for analyzing light-light-heavy-heavy (LLHH) correlators in conformal field theories. We specifically apply this formalism to correlators of giant gravitons in $\mathcal{N}=4$ super Yang-Mills to probe the nontrivial physics beyond planarity. By combining this framework with bootstrap techniques, we compute all four-point functions at strong coupling involving two maximal giant gravitons and two supergravitons of arbitrary dimensions. We identify a partially broken, higher-dimensional hidden symmetry -- a defect extension of 10d hidden conformal symmetry -- present at both strong and weak coupling, which allows these correlators to be packaged into a single generating function. Furthermore, we perform a systematic OPE analysis of the strong-coupling correlators, extracting the complete spectrum of anomalous dimensions for the defect-channel double-particle operators. Finally, we argue that the defect perspective provides the natural nonperturbative description for any LLHH correlator by showing that four-point conformal blocks reduce to defect two-point blocks in the heavy limit.

Using a data sample of 427.9 fb$^{-1}$ collected by the Belle~II detector at or near the $Υ(4S)$ and $Υ(10753)$ resonances, the cross sections for $e^+e^-\to h^+h^-J/ψ$ $(h=π/K/p)$ at center-of-mass energies ranging from 3.8 GeV or the production threshold to 5.5/6.0/7.0 GeV have been measured via initial-state radiation. The cross sections for the processes $e^+e^-\to π^+π^-J/ψ$ and $e^+e^-\to K^+K^-J/ψ$ are consistent with previously published results. The cross sections for these channels obtained by combining with previous Belle results are also given. The process $e^+e^-\to p\bar p J/ψ$ is investigated for the first time. The yields are small and no significant structure is observed in the cross section versus energy. Searches for vector charmonium-like states in the $h^+h^-J/ψ$ systems, and for associated intermediate states in the $h^{\pm} J/ψ$ systems, are also presented.

We revisit scalar $φ^4$ theory and construct a reorganized perturbative expansion in which the kinetic operator, rather than the quartic interaction, is treated as the perturbation. Starting from the exactly solvable $0$-dimensional model, we show that the resulting series is convergent for positive coupling and can be written as an expansion in negative powers of the quartic coupling $λ$. We extend the construction to higher-dimensional field theory using an auxiliary field, and we formulate a discrete lattice version in which multi-site contributions are systematically organized. We explicitly compute the leading terms in the expansion, study their continuum limit, and compare against brute-force numerical evaluations of the partition function. We discuss the relation of this expansion to standard weak-coupling perturbation theory, strong-coupling expansions, and resummation techniques, and we outline possible applications to nonperturbative studies of scalar field theories.

In this work, we investigate optical effects in modified Maxwell electrodynamics (ModMax) in the presence of an external electromagnetic field. Considering uniform and constant magnetic and electric backgrounds, the solutions for the refractive indices are revisited. Using these results, we obtain the propagating modes and the phase shift (birefringence) for plane wave solutions in the presence of a pure magnetic background field. Afterwards, we investigate the Goos-Hänchen effect considering the interface between a simple dielectric and a medium whose electromagnetic response tensors are ruled by the ModMax electrodynamics. Further, based on the general reflection problem, we discuss the complex Kerr rotation with both the electric $(E)$ and magnetic $(B)$ background fields, considering two main cases: i) $B>E$ and ii) $E>B$. Our findings indicate that the $γ$ parameter and ratios $(B/E)$ and $(E/B)$ play a central role in describing the Kerr signals (rotation and ellipticity) of systems with optical effects induced by non-linear electromagnetic effects.

We generalize prior work on the signatures of bulk signals detected by brane-based observers in a spacetime with a compactified dimension. When such braneworlds move with constant velocities or constant proper accelerations in the extra dimension, the observers may witness apparent superluminal signaling. Our analysis includes tilted branes that are partially wrapped along the compact dimension. We identify parameters that help characterize various scenarios. Despite the apparent superluminality of bulk signals, we show that anisotropies in their propagation relative to the brane-based observer preserve causality. Some of the effects studied here could be the basis for alternative cosmological models, as well as observable signatures of braneworld scenarios.

We construct new families of type-IIB supergravity solutions by employing TsT transformations on the ten-dimensional geometry that arises after the uplift of the five-dimensional soliton solution of Anabalón, Nastase, and Oyarzo. In particular, we identify two marginal and two dipole deformations of the uplifted geometry. We then analyse a plethora of holographic observables -- including Wilson loops, `t~Hooft loops, Page charges, entanglement entropy, and central charge -- and compare their behaviour across the different deformed backgrounds.

Dark sectors consisting of atomic constituents (electrons, protons, and photons) offer a well-motivated extension to the Standard Model while providing multiple avenues for phenomenological study. In this work, we explore the impact of conversions between the dark and Standard Model photons in the primordial CMB spectral distortion epoch ($10^3 \lesssim z \lesssim 10^6$). These conversions are resonantly enhanced when the induced thermal masses of both photonic species are equal, thus leading to the possibility that sizeable distortions can be produced. To this end, we solve the Boltzmann equation at early times to determine the (irreducible) freeze-in or freeze-out abundance of dark photons. This procedure also allows us to update the limits on generic milli-charged dark sectors using the ACT DR6 bound on the number of effective radiative degrees of freedom ($N_{\rm eff}$). By then modeling the evolution of the thermal masses in both sectors, we compute the primordial CMB distortion using the Landau-Zener formalism. We find that when the dark electron and proton are roughly similar in mass (the positronium limit), current spectral distortion data from the COBE/FIRAS instrument is able to rule out novel regions of parameter space. We also forecast bounds from the proposed FOSSIL satellite, finding that spectral distortions can also be used to probe the ultra-low dark electric charge regions of parameter space, which are difficult to investigate by other means.

We present new numerical results on the space of local, unitary, parity-preserving conformal field theories (CFTs) in three dimensions from the stress tensor bootstrap. In bounds maximizing certain OPE coefficients, we find a plethora of sharp features, such as kinks and ridges, as a function of scaling dimensions. We show that some of these features correspond to known theories, but there are many others that are equally strong but do not match known CFTs. We argue that these features are robust to raising numerical order and could then correspond to numerous as yet unknown CFTs. We conclude in proposing a program of "CFT cartography": the systematic exploration of the landscape of CFTs without individual theory targets in mind.

Long-period radio transients (LPTs) are a new class of radio sources characterized by long spin periods ($P_{\text{spin}}>10^3$ s) and highly variable radio emission. While known magnetars are relatively young ($τ<10^5$ yrs) with spin periods clustered between $1-10$ sec, it has been proposed that LPTs may be linked to a missing population of older magnetars. In this paper, we present an extensive parametric analysis of isolated magnetar spin evolution using various propeller spin-down models. In general, at higher initial magnetar B-fields ($B_0>\sim10^{15}$ G) and larger ambient densities ($n_0>\sim10^2$ cm$^{-3}$), magnetars will transition to the propeller phase earlier, and they start accreting gas from the ISM or molecular clouds after $τ\sim10^8$ yrs. We found that a transition from the pulsar to the propeller phase is required to reach the observed LPT period range of $P>10^3$ s. More specifically, our population synthesis study based on Monte-Carlo simulations shows that two propeller models can account for most of the observed LPT periods ($P\sim1-400$ [min]) and their period derivative constraints ($\dot{P}<10^{-9}$ s s$^{-1}$). Our spin-down models predict that (1) nearby radio-quiet neutron stars with the estimated dipole $B$-field range of $B\sim(1-5)\times10^{13}$ G will transition to the propeller phase eventually after $τ>\sim10^7$ yrs; (2) thermal X-ray emission from accretion-phase magnetars becomes too faint for detection after traveling ($d>\sim10$ kpc) from their birth places; (3) sporadic radio outbursts observed from LPTs may not be explained by regular radio pulsar and magnetar emission mechanisms that operate during the propeller phase.

Cross-survey generalization is a critical challenge in stellar spectral analysis, particularly in cases such as transferring from low- to moderate-resolution surveys. We investigate this problem using pre-trained models, focusing on simple neural networks such as multilayer perceptrons (MLPs), with a case study transferring from LAMOST low-resolution spectra (LRS) to DESI medium-resolution spectra (MRS). Specifically, we pre-train MLPs on either LRS or their embeddings and fine-tune them for application to DESI stellar spectra. We compare MLPs trained directly on spectra with those trained on embeddings derived from transformer-based models (self-supervised foundation models pre-trained for multiple downstream tasks). We also evaluate different fine-tuning strategies, including residual-head adapters, LoRA, and full fine-tuning. We find that MLPs pre-trained on LAMOST LRS achieve strong performance, even without fine-tuning, and that modest fine-tuning with DESI spectra further improves the results. For iron abundance, embeddings from a transformer-based model yield advantages in the metal-rich ([Fe/H] > -1.0) regime, but underperform in the metal-poor regime compared to MLPs trained directly on LRS. We also show that the optimal fine-tuning strategy depends on the specific stellar parameter under consideration. These results highlight that simple pre-trained MLPs can provide competitive cross-survey generalization, while the role of spectral foundation models for cross-survey stellar parameter estimation requires further exploration.

More than 150 measurements of the HeII 4686A emission line in spectra of SS433 were obtained during 388 nights in 2020-2025 with the Transient Double-beam Spectrograph on the 2.5 m telescope of Caucasian Mountain Observatory of Sternberg Astronomical Institute. We found that the HeII emission line formation region is not eclipsed and is significantly larger than both the donor star and the photosphere of the supercritical accretion disk. The HeII radial velocity curve was found to be independent of the precessional phase and inconsistent with the photometric curve. These findings suggest that the HeII line does not reflect the orbital motion of the compact object. Therefore, spectroscopic estimates of the masses of the components in SS433 based on the HeII emission line can be unrealistic.

The results of interferometric observations of the star-forming region S255IR in the frequency range 210--250 GHz are presented. The observations were carried out with the antenna array SMA (Hawaii, USA). Fifty-three molecules were detected, including complex organic molecules (COMs) such as CH$_3$CHO, CH$_3$CN, CH$_3$CH$_2$CN, and many others. Typical rotational temperatures in the hot core SMA1 fall in the range 100--200 K. Optical depths in the lines of methanol and some other molecules in the cores SMA1 and SMA2 were estimated. In SMA1, the optical depth of one of the strongest methanol lines, $5_{-1}-4_{-1}E$, proved to be $23.8 \pm 1.5$. Based on this value, one can assume that the lines of other oxygen-containing COMs, such as CH$_3$OCHO, CH$_3$OCH$_3$, CH$_3$CH$_2$OH, which are typically much less abundant in hot cores than methanol, are optically thin in SMA1. Most of the detected molecules can be roughly divided into two groups. The molecules of the first group emit exclusively toward the hot core SMA1, while some or all lines of the molecules of the second group, in addition to SMA1, can be seen toward a ring-like structure to the west of SMA1. This structure is most likely associated with the walls of a cavity formed by high-velocity outflows driven by young stellar objects (YSOs) in molecular cores SMA1, SMA2, and possibly SMA3. The gas temperature and density in the cavity walls were estimated using methanol lines. The temperature was found to be about 50--60 K, and the density about $10^7-10^8$ cm$^{-3}$. The column density of methanol near the brightness peaks in the lines of this molecule is about $5\times 10^{15}$~cm$^{-2}$. The column densities of other COMs in the ring-like structure will be determined in future studies with increased sensitivity achieved by spectral line stacking.

The [C/N]-age relation has become a powerful tool for reconstructing the formation history of the Milky Way (MW), providing the largest age sample for field giant stars. However, at metallicities below [Fe/H] $< -0.5$, stellar surfaces are altered by a poorly understood process known as extra mixing, which modifies [C/N] in a mass- and metallicity-dependent manner. This effect complicates the application of the traditional [C/N]-age relation in metal-poor regimes. Within the MW, constraining the mass dependence of extra mixing is particularly challenging because stars at [Fe/H] $< -0.5$ are predominantly old and therefore low-mass, leading to strong degeneracies between mass and metallicity. In this work, we explore the potential of the Magellanic Clouds (MCs) to disentangle these effects and constrain extra mixing as a function of age and metallicity. By comparing empirical corrections calibrated in the MW with predictions from thermohaline mixing models, we isolate the mass dependence of extra mixing in the MCs down to [Fe/H] $\sim-0.7$. We find that the empirical calibration performs well for lower-mass stars ($< 1.25$ $M_{\odot}$), while theoretical models successfully reproduce the observed mass dependence down to $\sim$ 1.25 $M_{\odot}$. We further present the first observational evidence that extra mixing becomes ineffective above $\sim$ 1.8 $M_{\odot}$ at [Fe/H] $\sim -0.7$. Our results demonstrate the feasibility of deriving [C/N]-based ages for individual stars in external galaxies. Future observations targeting higher-$\log g$ or fainter stars in the MCs will provide stronger constraints on extra-mixing processes and enable the calibration of [C/N]-age relation that can be applied to low-metallicity individual stars in the MW or external galaxies.

We present imaging and spectroscopic analyses of Chandra and XMM-Newton observations of ACT-CL J0123.5$-$0428, one of the most massive, highest redshift galaxy clusters detected within the survey fields of the Atacama Cosmology Telescope. The Chandra data are sufficient to characterize the morphology of this cluster and constrain the geometrically deprojected temperature in 2 spatial bins out to $r_{2500}$, revealing a dynamically relaxed system whose temperature drops to $kT = 1.8\pm0.6$ keV in the inner $\sim40$kpc. Within this same inner radius, the surface brightness and density of the ICM is sharply peaked, and the cooling time falls to $t_\mathrm{cool}=280^{+150}_{-120}$ Myr. A novel forward-modeling analysis of the XMM data extends imaging and spectroscopic measurements of this system out to $r_{500}$, constraining the redshift to $z=1.50\pm0.03$, with a mean temperature of $kT = 7.3\pm1.1$ keV and an emission-weighted mean metallicity of $Z/Z_\odot = 0.43^{+0.46}_{-0.25}$. We also utilize the limited optical/IR photometric coverage of the cluster to characterize the properties of the brightest cluster galaxy (BCG), which is coincident with the X-ray peak. Despite the high redshift and strong cool core, the BCG exhibits no signs of recent or ongoing star formation, suggesting AGN feedback has been acting persistently to stem star formation since $z\sim 2.5$. These measurements identify ACT-CL J0123.5$-$0428 as the highest redshift, dynamically relaxed, cool core galaxy cluster discovered to date, making it a premier target for future astrophysical and cosmological studies.

NGC 4631 X-8 is an ultraluminous X-ray pulsar (ULXP) having a spin period of about 9.7 s, discovered using XMM-Newton observations in 2025. The pulsar is known to show one of the largest spin-up rates ($\sim 9.6 \times 10^{-8}$ s s$^{-1}$) among the ULXP population. We explore the surface magnetic field of the neutron star in this source using different models, and find that the inferred magnetic field lies in the range of about $0.3-2 \times 10^{14}$G. We study the long-term magnetic field and spin period evolution of the pulsar assuming steady accretion using prevalent theoretical mechanisms and find that the pulsar will evolve to become a millisecond pulsar having decayed magnetic field of about $\sim 10^{9}$G in about a million years. The scenario of the formation of a millisecond pulsar is also probed using an estimate of the super-Eddington duty cycle of about 14% from the literature, which suggests that the neutron star would accrete sufficient matter to become a recycled millisecond pulsar. Exploring the magnetic field as well as the spin period evolution properties of ULXPs may enable us to understand the poorly understood evolutionary features of ULXPs, shed light on one of the pathways of millisecond pulsar formation and also help us to understand transient super-Eddington accretion phases in newborn magnetars, which are believed to power energetic events such as long gamma-ray bursts and Type I superluminous supernovae.

We investigate the bar fraction in galaxy pairs from the SDSS to assess how galaxy interactions affect bar structures. Compared to isolated galaxies, close pairs exhibit a significantly reduced bar fraction at projected separations within 25 kpc. This reduction is driven almost entirely by systems showing clear merger or disturbance signatures, indicating that tidal interactions suppress bars. The decline is dominated by a decrease in weak bars, while the fraction of strong bars remains largely unchanged. Bar suppression is primarily associated with major mergers and is strongest in massive host galaxies. A weaker but statistically significant suppression is detected in minor mergers only for massive galaxies with small bulges. In contrast, no significant dependence of bar suppression on the relative orientation between pair members is found. These findings provide observational evidence that tidal perturbations in major mergers play a key role in regulating bar evolution.

It has long been suspected that black hole accretion-outflow coupling is invariant from the stellar to supermassive scales. Stellar mass black hole accretion flows are known to launch jets and outflows as they transition through critical accretion rate thresholds, with values well constrained observationally owing to their short evolutionary timescales. In contrast, accretion flows in typical supermassive black hole (SMBH) systems (those in active galactic nuclei) evolve over thousands of years, making the critical transitions at which jets are launched impossible to constrain in individual systems. Tidal disruption events (TDEs) provide the unique opportunity to witness the birth and evolution of an accretion flow onto a SMBH which evolves on timescales of years. Here we show that TDEs launch outflows during a super-Eddington accretion phase and a second, physically distinct outflow, at a critical accretion rate of $L_{\rm crit} \approx0.02$ $L_{\rm Edd}$, the same as the critical accretion rate for state transitions observed in accreting stellar mass black holes. This work naturally explains the mechanism, observed properties, and detection rate for prompt and delayed outflows observed in TDEs, which until now have been open problems. More broadly, we demonstrate that SMBHs exhibit the same accretion-outflow coupling as stellar mass black holes and that the critical low accretion rate threshold for jet formation in black holes is scale invariant.

The angular, harmonic cross-correlation between gravitational wave (GW) events and galaxy catalogues contains rich information on the large-scale structure and the origin of compact binary mergers. In this work, we study how uncertainties in the binary formation channel affect the predicted cross-correlation signal for both current-generation and next-generation networks of detectors. We generate five mock GW catalogues for which we vary the progenitor-to-remnant mass-transfer function and the time-delay probability distribution between progenitor and remnant. We then cross-correlate these catalogues with galaxy samples modelled on the 2MASS Photometric Redshift catalogue (2MPZ) and the Gaia-unWISE quasar catalogue (Quaia). We find that the mass-transfer function has negligible effect on the cross-correlation signal, with differences remaining within redshift uncertainties. In contrast, the time-delay distribution dramatically affects the redshift distribution of the GW events and, with it, the cross-correlation signal, particularly for shallow galaxy catalogues. In particular, current-generation facilities can achieve significant detections only for the longest time delays when cross-correlated with 2MPZ, whilst all cross-correlations with the deeper Quaia catalogue are marginally detectable or consistent with zero. Our exploratory results thus demonstrate that forecasts on cosmological or astrophysical parameters derived from GW-galaxy cross-correlations are, as expected, strongly sensitive to the assumed binary formation history.

Astrochemical observations have revealed a surprisingly high level of chemical complexity, including long carbon chains, in the earliest stages of Sun-like star formation. The origin of these species and whether they undergo further growth, possibly contributing to the molecular complexity of planetary systems, remain open questions. We present recent observations performed using the 100-m Green Bank Telescope of the prestellar core L1544 and the protostellar system IRAS 16293-2422. In L1544, we detected several complex carbon-bearing species, including $\mathrm{C_2S}$, $\mathrm{C_3S}$, $\mathrm{C_3N}$, $\mathrm{c\text{-}C_3H}$, $\mathrm{C_4H}$, and $\mathrm{C_6H}$, complementing previously reported emission of cyanopolyynes. In IRAS 16293-2422, we detected $\mathrm{c\text{-}C_3H}$ and, for the first time, $\mathrm{HC_7N}$. Thanks to the high spectral resolution, we refine the rest frequencies of several $\mathrm{c\text{-}C_3H}$ and $\mathrm{C_6H}$ transitions. We perform radiative transfer analysis, highlighting a chemical difference between the two sources: IRAS 16293-2422 shows column densities 10-100 times lower than L1544. We perform astrochemical modeling, employing an up-to-date chemical network with revised reaction rates. The models reproduce the general trends, with cyanopolyyne and polyynyl radical abundances decreasing as molecular size increases, but they underestimate the abundances of cyanopolyynes longer than $\mathrm{HC_5N}$ by up to two orders of magnitude. Current models, which include the dominant neutral-neutral formation routes, cannot account for this discrepancy, suggesting that the chemical network is incomplete. We propose that additional ion-molecule reactions are crucial for the formation of these species. Developing a more comprehensive chemical network for long carbon chains is essential for accurately interpreting present and future observations.

Sensitivity of searches for extended emission around gamma-ray sources is naturally limited by the precision of the knowledge of the Point Spread Function (PSF) of gamma-ray telescopes. Inaccuracies in the PSF models of the Fermi Large Area Telescope (LAT) can potentially lead to false positive detections of source extension. We explore uncertainties in the Fermi/LAT PSF by comparing the PSF models provided by the Fermi/LAT Instrument Response Functions (IRFs) with signals of bright pulsars. We compare the analytical PSF models of Fermi/LAT IRFs with pulsar data and fit the pulsar data with the same analytical model as in the Fermi/LAT IRFs to derive an improved set of PSF parameters. We then apply this revised PSF parameterisation to the search of extended emission around a blazar, Mrk 501. We find that the parameters of the analytical PSF models of Fermi/LAT IRFs are inconsistent with the pulsar data. We obtain an improved set of PSF parameters from the fits to pulsar data that is consistent with observations. We find no evidence of the previously reported extended signal around Mrk 501 if the revised PSF consistent with pulsar data is used in data analysis.

We investigate whether a non-minimal coupling between dark matter and gravity can influence the internal dynamics of ultra-diffuse galaxies. Within this framework, the gravitational potential is modified by an additional term that captures the interaction between spacetime curvature and the dark matter with a coupling constant determined by a length scale $L$. Using spherical Jeans modeling, we analyze the kinematic data of three ultra-diffuse galaxies namely: NGC\;1052-DF2, NGC\;1052-DF4, and Dragonfly\;44, which span the observational extremes from dark matter deficient to dark matter dominated systems. For each galaxy we explore several dark matter halo profiles, two orbital anisotropy models, and both with and without Stellar-to-Halo Mass Relation scenarios, and we perform a Bayesian parameter inference. Our results show that across all the considered configurations, the constrained astrophysical parameters are consistent with standard ones from General Relativity. The posterior distributions of $L$ show no preference for non-zero values and result only in upper limits, suggesting that any non-minimal coupling contribution must be small and perturbative on this scales. Future high precision velocity measurements will be essential to determine whether non-minimal coupling effects can become observationally distinguishable in low-acceleration systems.

Quantum gravity theories often modify spacetime symmetries. In particular, Lorentz invariance may be violated when approaching the Planck scale. Although the scales at which interactions occur in extensive air showers induced by ultra-high-energy cosmic rays in the atmosphere are many orders of magnitude below the Planck scale, these violations might still be observable. In this work, the fluctuations in the number of muons in the extensive air showers measured at the Pierre Auger Observatory are exploited, for the first time, to constrain Lorentz invariance violations. The bounds derived in the hadronic sector are the strongest ever obtained, and do not rely on assumptions about the mass composition of ultra-high-energy cosmic rays. The fluctuations in the number of muons constitute a new and powerful observable to further explore Lorentz invariance in a region of the parameter space not accessible to other observables.

4U 1556-60 is a low-mass X-ray binary that was discovered more than 50 years ago as a persistent X-ray source; however, very little was known about it. Recently, Gaia obtained a parallax for the optical counterpart that places 4U 1556-60 at a distance of only about 700 pc, making it one of the closest X-ray binaries known to date. This close distance drastically alters what was previously assumed about the source. We revisit 4U 1556-60 in light of the newly determined distance of 700 pc, reinterpreting its literature and presenting new X-ray and radio observations to better understand various characteristics of the system. We conclude that a scenario in which 4U 1556-60 is a candidate ultracompact neutron star X-ray binary at a distance of ~700 pc is able to explain the observed properties of the source. It resides at a persistent X-ray luminosity of ~2x10^34 erg/s, an unusual value for a typical X-ray binary, but similar to several ultracompact systems. The ratio of the X-ray to optical luminosity is very high, also suggesting a physically small accretion disk. The radio jet is undetected with a very deep upper limit of 3x10^25 erg/s, which is about 10^3 times fainter than the expected black hole jet correlation, strongly indicating a neutron star accretor. The X-ray spectrum is dominated by a power law, and the X-ray timing properties are also consistent with observations of other very low accretion rate X-ray binaries. No spin or orbital periodicity are found in the X-ray data. Future observations, especially to determine its orbital period, will further aid in understanding 4U 1556-60.

Gravitational-wave observations of binary neutron star systems can shed light on the currently unknown dense matter equation of state. The equation of state determines a large number of neutron star properties, such as tidal deformability, radius, and quadrupole moment, several of which directly affect the emitted gravitational-wave signals. To reduce the dimensionality when computing gravitational-waves and when interpreting observational data, quasi-universal relations are commonly employed to connect different neutron star properties. However, quasi-universal relations are not exact and their use may introduce uncertainty and bias. We explore the potential biases arising from different quasi-universal relations in the third generation era: (i) the Love-Q relation connecting the spin-induced quadrupole moment and the tidal deformability, (ii) the relation between the fundamental mode frequency and the tidal deformability, and (iii) the binary Love relation. We find that for the quadrupole relation biases are only present for rapidly rotating systems, for the binary-Love relation induces moderate biases only in the next-to-leading-order tidal parameters, which can however propagate into the inferred equation of state at low masses. Regarding fundamental mode frequencies, we find that the employed relation introduces only negligible biases, while waveform systematic effects can become comparatively large. Our results highlight that while quasi-universal relations remain a useful tool within gravitational-wave analyses, careful treatment is needed to avoid biases in equation of state measurements with next-generation detectors.

HESS J0852-463 is a TeV γ-ray source located in the Galactic plane. The region consists of a supernova remnant (SNR, RX J0852.0-4622) with a shell-like morphology, commonly referred to as Vela Junior, and a pulsar denoted PSR J0855-4644. Pulsars are among the most efficient leptonic accelerators in our Galaxy, making this region particularly interesting to study. We utilise the most recent data taken by the High Energy Stereoscopic System (H.E.S.S.), to investigate any γ-ray emission associated with the pulsar in this region, PSR J0855-4644. We applied a full forward folding method on the H.E.S.S. data. Utilising 3D modelling techniques, we evaluated the TeV γ-ray emission towards the various components of this complex system. The distinct energy-dependent morphology observed in our data motivates further investigation of this source. We resolved the emission in the Vela Junior region into various components, several of which correspond to the SNR itself. In particular, we find a new extended component which is coincident with the position of PSR J0855-4644. The spectrum follows a power-law distribution with a best-fit index of ΓE = 1.81 \pm 0.07stat which differs from the properties of the surrounding γ-ray emission of the Vela Junior SNR. A one-zone leptonic joint fit between the X-rays (from XMM-Newton) and γ-rays (from H.E.S.S.) leads to a lower limit on the magnetic field of 1.6μG and a spectral index of α = 1.88 \pm 0.01, in line with expectations of pulsar wind nebulae (PWNe). In this paper, we report the first detection of the PWN of PSR J0855-4644 at TeV energies with the H.E.S.S. experiment at a significance of 12.2σ. This is attributed to the advanced techniques of the 3D analysis. Based on the pulsar's characteristics, its PWN is consistent with the known TeV PWNe population in the Galaxy.

Cosmic voids contain higher-order cosmological information and are of interest for astroparticle physics. Finding genuine matter underdensities in sparse galaxy surveys is, however, an underconstrained problem. Traditional void finding algorithms produce deterministic void catalogs, neglecting the probabilistic nature of the problem. We present a method to sample from the stochastic mapping from galaxy catalogs to arbitrary void definitions. Our algorithm uses a deep graph neural network to evolve "test particles" according to a flow-matching objective. We demonstrate the method in a simplified example setting but outline steps to generalize it towards practically usable void finders. Trained on a deterministic teacher, the model performs well but has considerable stochasticity which we interpret as regularization. Cosmological information in the predicted void catalogs outperforms the teacher. On the one hand, our method can cheaply emulate existing void finders with apparently useful regularization. More importantly, it also allows us to find the Bayes-optimal mapping between observed galaxies and any void definition. This includes definitions operating at the level of simulated matter density and velocity fields.

To explore the environmental mechanisms causing quenching in nearby star-forming galaxies, we study the variation with local and large-scale environments of a star formation concentration index, C-index $\equiv\log{(r_{50,{\rm H}α}/r_{50,\rm cont}})$, that traces the spatially-resolved distribution of H$α$ emission. Our analysis combines (i) GAMA spectroscopic redshift survey data to optically select galaxy groups and reconstruct the cosmic web, (ii) eROSITA data to identify X-ray-emitting groups, and (iii) SAMI Galaxy Survey data to characterise spatially-resolved star formation. We find that galaxies in X-ray+optical groups exhibit the lowest median C-index and the highest fraction of centrally-concentrated star-forming galaxies relative to optical groups and the field (independently of group or stellar mass). Star-forming galaxies in more X-ray luminous groups at fixed dynamical mass show more concentrated star formation. At large scales, nodes show the lowest median C-index and the highest fraction of centrally-concentrated star-forming galaxies relative to filaments and voids, which have similar C-index distributions. C-index correlates most strongly with the distance to the closest node, leaving no significant role for other local or large-scale environment metrics. Finally, regular star-forming galaxies tend to have spins aligned parallel to filaments, consistent with smooth gas accretion, while centrally-concentrated galaxies tend have spins aligned perpendicular to filaments, likely driven by mergers and associated with bulge growth. These results suggest that multi-scale environmental processes, i.e. locally and at large-scale, act to concentrate star formation toward galaxy centres, via gas-related mechanisms in nodes and ram-pressure stripping in X-ray+optical groups.

Context. Solid methane (CH$_4$) is an important molecule in interstellar and planetary environments, serving as a precursor to complex organic compounds and a potential biosignature in exoplanetary studies. Despite its significance, laboratory data on low-temperature phase of methane below 10 K remain limited. Aims. We aim to obtain spectra of methane in binary mixtures at 10 K and compare it to the spectra obtained at 6.7 K. These temperatures correspond to phases II and II* of pure methane and are representative of dark molecular clouds and protostars at early stages. We also aim to test the obtained data applicability to JWST data interpretation. Methods. Laboratory reference spectra were obtained on the ISEAge setup via FTIR spectroscopy in transmission mode. A weighted $χ^2$ minimization is used for the fitting. Results. We present infrared spectra with corresponding band strengths of pure methane and binary mixtures with methane: CH$_4$:H$_2$O,CH$_4$:CO$_2$, CH$_4$:CH$_3$OH, CH$_4$:NH$_3$ at 6.7 K and 10 K showing a 20\% increase in mixtures compared to commonly used 10 K band strength value of pure methane. We also test the usability of the spectra on open JWST data by probing the spatial distribution of methane in B335. We also present additional experiments concerning the phase transition of methane between phase II* and phase II. Conclusions. Our results reveal distinct spectral features for methane in non-H$_2$O environments, enabling more accurate interpretation of JWST observations. The dataset of spectra, publicly available on Zenodo, can be used for fitting JWST data.

In a recent arXiv posting [1] I reported the result of an experiment: asking Perplexity.ai to compare three items concerning (ordinary) Gamma Ray Burts (GRBs): the data, the standard paradigm(s) and the "Cannonball" (CB) model. Here I ask the same URL to extend this comparison to long--lasting GRBs, binary Neutron-Star mergers and their associated short--hard GRBs, low--luminosity GRBs, X--ray flashes, X--ray transients, and non--solar cosmic rays. The results of this experiment are enlightening but worrisome. Except for this abstract, two footnotes and two other references to standard [2] and CB-model [3] articles and talks, all of what follows is, verbatim, what the cited AI "opines".

Foreground masking and incomplete sky coverage complicate CMB polarization analyses by inducing mode coupling and imperfect E/B separation, with particularly strong impact on searches for primordial $B$-modes. We present SkyReconNet-P, a convolutional neural network for inpainting CMB polarization maps that extends the SkyReconNet framework to jointly reconstruct the polarization $(Q,U)$ maps from partial-sky observations. The method combines regional processing with a hybrid design, utilizing standard convolution and dilated convolution to do a multiscale feature integration. We evaluate performance at both the map and power spectrum level using two masking scenarios: a generated random mask and the Planck 2018 common polarization inpainting mask. For both masking scenarios, SkyReconNet-P reproduces the large-scale morphology of the target maps. In power-spectrum space, we find that the reconstructed $E$-mode spectrum closely tracks the target at low multipoles, while small biases emerge at higher $\ell$. For $B$-mode, the raw reconstructed spectra exhibit a larger multipole-dependent bias, which we mitigate using a simulation-based linear calibration. We show that the calibrated $B$-mode spectrum preserve more information by comparing it with spectrum estimation using pseudo-$C_\ell$. Finally, we demonstrate cosmological parameter inference from calibrated reconstructed spectra by fitting $(r, A_{\rm lens})$ with a Gaussian bandpower likelihood, recovering posteriors consistent with injected parameters across three test ensembles down to $r \sim 10^{-3}$. These results support inpainting as a complementary route to cut-sky approaches when downstream pipelines can greatly benefit from statistically well-characterized, gap-filled polarization maps.

A model-agnostic description of the baryon acoustic oscillation (BAO) feature in redshift space requires a number of ingredients. Physically, one must describe the impact of cosmological bulk flows which progressively and anisotropically smear out the feature over time. One must also model the effects of the scale dependence of tracer bias and the mode coupling between short and long scales. All of these can be incorporated using the Zel'dovich approximation alone, without reference to any particular cosmological model. On the technical front, one needs a robust, complete and cosmology-independent basis to describe the shape of the real space BAO feature in linear theory, which can then be propagated to the nonlinearly evolved, measured feature in redshift space. In this work, we describe how these ingredients -- which we have systematically constructed in recent work -- come together in an accurate framework capable of describing the BAO-scale pairwise measurements of state-of-the-art galaxy surveys. Using mock observations and $N$-body simulations, we show that our template-free framework can produce unbiased and precise cosmological constraints for samples with realistic levels of nonlinearity. This work represents one of the final steps in constructing a data-ready analysis framework for model-agnostic cosmological inference from the BAO feature.

We report the Fe K relativistic reverberation feature for the first time in the Narrow-line Seyfert\,1 galaxy Ton\,S180. Using a long observation from {\it XMM-Newton} we find that the Fe K emission lag peaks at $117\pm49$ s in the lag energy spectrum computed for frequencies $(0.3-8.5) \times 10^{-4}$ Hz. The lag amplitude drops to $22.85\pm14.20$ s as the frequency increases to $(8.5-30) \times 10^{-4}$ Hz. The time-averaged spectrum of the source shows a relatively narrow Fe K line at $\sim6.4$ keV, resulting in black hole spin to be low ($a=\rm 0.43_{-0.14}^{+0.10}$) found from the reflection modelling. We perform general relativistic transfer function modelling of the lag energy spectra individually. This provides an independent timing-based measure of the spin at $a=0.30_{-0.17}^{+0.34}$, and black hole mass $M_{\rm BH} = 0.29_{-0.16}^{+0.01}\times10^8M_{\odot}$, comparable to the previous measurement, and height of the corona $h = 2.59_{-0.33}^{+5.17}r_{\rm g}$. Further, we observe that the Fe K lag and the black hole mass fit well in the linear lag-mass relation shown by other Seyfert 1 galaxies.

We develop a self-consistent nonlinear extension of diffusive shock acceleration that incorporates cosmic ray (CR) backreaction on the shock precursor together with a physically motivated upstream-escape mechanism that produces an exponential high energy cutoff. The CR pressure gradient decelerates the upstream flow facing the shock wave, generating an extended precursor in which higher rigidity particles sample a larger cumulative velocity gradient and thereby acquire a progressively harder spectrum. Finite-size/escape effects are modeled by a momentum-dependent loss term, which naturally terminates acceleration and steepens the spectrum near the cutoff. The precursor compression ratio is not imposed as a closure condition but is determined dynamically by requiring consistency between the injection rate inferred from thermal leakage at the subshock and the injection strength demanded by the nonlinear shock modification, with CR-driven wave heating providing stabilizing negative feedback. Applying the model to young supernova-remnant-like parameters and standard one-zone Galactic diffusion, we reproduce the main features of the latest DAMPE proton spectrum: gradual hardening from hundreds of GeV to multi-TeV energies and a subsequent exponential cutoff at tens of TeV. The resulting spectral evolution follows directly from the competition between precursor-mediated nonlinear feedback and upstream escape.

We present a stellar stream found in images of the nearby, nearly face-on, late-type galaxy, NGC 3938 obtained for the LBT Imaging of Galactic Halos and Tidal Structures (LIGHTS) survey that is thin, has very low mean surface brightness ($\langleμ_g\rangle \approx$ 28.7 mag arcsec$^{-2}$ and $\langleμ_r\rangle \approx$ 28.1 mag arcsec$^{-2}$), appears to lie nearly on the plane of the sky, and wraps more than half way around a host galaxy that is otherwise apparently isolated. We estimate that the progenitor had a stellar mass of $\sim 3.7\times 10^7$ M$_\odot$. Despite an intriguing apparent offset between the centroid of the host galaxy and the apparent center of the stream orbit, we find that we can reproduce the morphology, including this apparent off-centering, with simple models and standard assumptions about the host (thin disk centered within a canonical spherical dark matter halo) and the progenitor satellite orbit. We identify a number of detailed features of the stream, such as changes in curvature and density, that will require more complex models to reproduce. Even this rather simple system provides a rich set of constraints with which to explore the accretion history and gravitational potential of an otherwise unremarkable late-type galaxy. Given the depth of the LIGHTS images, this system is an example of the types of stellar stream that could be found in a majority of nearby giant galaxies with the 10-year stack of Rubin/LSST data.

The center of the Galaxy is a prominent source in X-rays and gamma-rays. The study of its high-energy (HE) emission is crucial in understanding the physical phenomena taking place in this dense and extreme environment, where the closest supermassive black hole (SMBH) to us, Sgr A*, is lurking nearly invisible, today, in most of the energy spectrum. These phenomena are probably common to other galactic nuclei and may explain the feedback processes between nuclear regions and galaxies, so important for the overall evolution of the Universe. The Galactic center HE emission is very complex and consists of both thermal and non thermal radiation produced by compact and extended sources, surrounded by more diffuse components. All these objects and media are interacting with each other in the narrow and dense Central Molecular Zone (CMZ). Some of them also show relevant extensions towards the Galactic poles, indicating energetic outflows that seem to link the center to the recently observed large Galactic polar structures. In spite of the fundamental advances obtained in the last twenty five years with the most sensitive X-ray and gamma-ray observatories, several questions remain open to investigations. We review here the main observational results and the open issues on the high-energy diagnostics of the Galactic nuclear activity, focusing on processes that take place in the CMZ, and in particular discussing the role of the present and past SMBH activities in powering this region and possibly the whole Galaxy.

We investigate the potential birthplace of Galactic binary neutron star (BNS) systems through a kinematic analysis. Using high-precision astrometry from Gaia DR3, updated pulsar distances, and Monte Carlo sampling of astrometric errors, we integrate the past trajectories of 11 Galactic BNSs and 167 globular clusters plus 2967 open clusters, to search for past encounters. Our results suggest that BNS origin in globular clusters is unlikely, with low encounter probabilities (e.g., $\lesssim 0.5\%$ for NGC 5139) and requiring excessive ejection velocities. Conversely, our analysis indicates that open clusters are a non-negligible formation channel. Specifically, the double pulsar J0737$-$3039 shows a $13.9\%$ ($5.4\%$) probability of originating from the young cluster OC 0450 (Theia 58). Based on encounter proximity and time, we argue that Theia 58 is its more plausible birthplace. Our work provides kinematic evidence consistent with an open-cluster origin for a subset of field BNSs.

Magnetic properties of dust grains play a fundamental role in their alignment with ambient magnetic fields and magnetic dipole emission. In the radiative torque (RAT) paradigm, superparamagnetic inclusions (SPIs) embedded within dust grains are expected to significantly enhance magnetic susceptibility and alignment efficiency. Previous studies have generally assumed SPIs of a single characteristic size. In this work, we develop an effective superparamagnetism model that explicitly accounts for a power-law size distribution of SPIs. We show that the effective zero-frequency susceptibility can be described by the superparamagnetic susceptibility of uniform-sized inclusions evaluated at the critical blocking size, reduced by a factor $F_{\rm eff}\sim 0.1$. It exhibits a slight increase with dust temperature $T_{d}$, in contrast to the rapid decrease for the case of single-size SPIs. For rotating grains at angular frequency $ω$, we identify a characteristic resonance size of SPIs that dominates the magnetic response, $N_{\rm res} = (T_{d}/T_{\rm act}) \ln (ν_{0}/ω)$ with $T_{\rm act}$ activation temperature and $ν_{0}$ the characteristic attempt frequency of SPIs. The frequency-dependent effective susceptibility is well described by the maximum susceptibility $χ_{\rm eff}^{\rm max}(ω)$ at $N_{\rm res}$, reduced by a factor $G_{\rm eff}\sim 0.1$. Unlike models assuming uniform-sized inclusions, we find that the effective susceptibility exhibits a nearly flat spectrum for frequency below $ν_{0}$, arising from the progressive activation of larger inclusions at lower frequencies. This effective superparamagnetism model based on the SPI size distrbution has important implications for magnetic grain alignment, dust polarization, and magnetic dipole emission across diverse environments.

We combine narrow/medium-band filter photometry from the Southern Photometric Local Universe Survey (S-PLUS) DR4 with ultra broad-band filter photometry from Gaia EDR3 to derive fundamental stellar parameters ($T_{\rm eff}$, $\log g$, [Fe/H], ages) and elemental-abundance ratios ([C/Fe] and [$α$/Fe]) for 5.4 million stars in the Galaxy (4.9 million dwarfs and 0.5 million giants), as well as for over 0.7 million red giant stars in the Large and Small Magellanic Clouds (LMC and SMC). The precisions of the abundance estimates range from 0.05-0.10 dex for metallicity in the relatively metal-rich range ([Fe/H] $> -1.0$) to 0.10-0.30 dex in the metal-poor regime ([Fe/H] $<-1.0$), 0.10-0.20\,dex for [C/Fe], and 0.05 dex for [$α$/Fe]. The stellar parameters for LMC and SMC member stars are somewhat less precise than those from the S-PLUS main survey, primarily because of the effect of high reddening. The use of both metallicity- and carbon-sensitive filters provides unbiased measurements of both [Fe/H] and [C/Fe], of particular importance for very low-metallicity ([Fe/H] $< -2.0$) stars, where carbon enhancement can lead to systematically high estimates of [Fe/H] when only a single metallicity-sensitive filter is employed. Furthermore, multiple narrow-band filters enable metallicity estimates down to [Fe/H] $\sim -4.0$ with an accuracy of around 0.3 dex, exceeding the precision typically achieved by low/medium-resolution spectroscopy. This extensive photometric dataset, combined with the other three datasets in this series, will serve as a valuable legacy resource for Milky Way and Magellanic Clouds studies.

We present a parametric strong-lensing model for the galaxy cluster MACS J1931.8-2635 ($z_l = 0.35$), accompanying the detection of the spectroscopically confirmed SN Eos at $z = 5.13$ (Coulter et al. 2026). We identify 10 new multiple-image systems in recent VENUS JWST/NIRCam imaging, so that the model is constrained with a total of 19 robust multiple-image systems -- nine of which also have a spectroscopic redshift. For the point-like source corresponding to SN Eos, our model predicts a total of five images, with the observed radial image pair having a similar magnification of $μ\simeq 25 - 30$ and a small time delay of $< 5$ days, in agreement with their simultaneous observation. According to the model, the other three predicted images arrived earlier, with time delays of $3.7 \pm 0.7$, $3.5 \pm 0.7$ and $54.0 \pm 10.8$ years prior to the two observed images, and with magnifications of $12.9 \pm 2.6$, $13.0 \pm 2.9$ and $2.2 \pm 0.4$, respectively. The absence of detections at the predicted positions, where the host galaxy's images are also visible, confirms the transient nature of the source. SN Eos and its host galaxy are studied in separate articles, and we here focus on the lens model. The final model reaches a very good $r.m.s.$ distance between model and observations of $0.44''$. We present the lens-modeling results, including newly identified systems such as a triply imaged, grand-design spiral galaxy candidate at $z \simeq 3.65_{-0.09}^{+0.04}$, and briefly discuss the potential of using high-redshift lensed SNe for cosmography.

Lorentz symmetry is a fundamental pillar of modern Physics, yet high-energy theories often predict its violation. One potential signature of such a violation is cosmic birefringence - rotation of the polarization plane of photons due to Chern-Simons coupling in Maxwell's electrodynamics. This rotation angle, aka birefringence angle, depends upon the distance travelled by the photon and is thus different for CMB and 21cm photons. While the rotation angle in CMB, i.e., $α_\mathrm{CMB}$, has been tightly constrained by CMB experiments, the potential of the 21cm cosmological signal to constrain this parameter, as well as constrain $α_\mathrm{21cm}$, remains largely unexplored. In this work, we provide constraints on both these angles by cross-correlating 21cm and CMB signals. Using the Fisher matrix formalism, we give our forecasts for 21cm experiments, including SKA, HIRAX, and PUMA, and Planck like CMB experiment. We find that best constraints $σ_{α_\mathrm{CMB}} \sim 4.4^\circ$ and $σ_{α_\mathrm{21cm}} \sim 100^\circ$ are found using $C_\ell^{T_{21} B_\mathrm{CMB}}$ and $C_\ell^{T_{21} B_{21}}$ respectively. Since birefringence hasn't yet been detected in 21cm, we choose the fiducial value $α_\mathrm{21cm}^\mathrm{fid}=0$ assuming the null hypothesis.

The past decade has been a transformative period for pulsar timing arrays (PTAs) and their search for nanohertz gravitational waves (GWs). This progress has been driven by collective advances in instrumentation for pulsar timing observations, increasingly sophisticated data-analysis techniques, and improved theoretical understanding of the origins of nanohertz GW signals. PTA sensitivity has steadily improved, leading first to progressively more stringent upper limits on the gravitational-wave background (GWB), and subsequently to the identification of a common red-noise process in pulsar timing data, the first hint of a GWB. In 2023, multiple PTA collaborations reported evidence for the Hellings-Downs correlation, widely regarded as the definitive signature of a GWB. These developments place PTAs on the threshold of a confident GW detection and the opening of a new low-frequency window on the GW Universe. In this article, we present an overview of PTA experiments, with particular emphasis on the rapid progress achieved during this pivotal period for PTA and nanohertz GW science.

An analysis of the Millimetre Astronomy Legacy Team 90 GHz (MALT90) survey has produced a sample of 27 candidate dense molecular clumps with large collapse motions, as revealed by large ``blue'' asymmetrical line profiles of the optically thick \hcop\, line. %with respect to the optically thin \nthp\, line. New, more sensitive molecular line observations of this sample, conducted with the Mopra 22-m telescope, confirm the blue asymmetries in the \hcop\, line profiles, with large, positive values of the asymmetry parameter $A$ ($\bar{A}_{HCO^+} = 0.69\pm0.01$), and positive, but smaller asymmetries in the \hcn\, and \hnc\, lines: ($\bar{A}_{HCN} = 0.35\pm0.01$ and $\bar{A}_{HNC} = 0.28\pm0.01$), as expected for a less optically thick tracer in collapsing clumps. The small, positive mean asymmetry parameters for \cch\, and \htcop, $\bar{A}_{C_2H} = 0.15\pm0.02$ and $\bar{A}_{H^{13}CO^+} = 0.18\pm0.03$, likely indicate slightly optically thick emission for at least some clumps. The hyperfine ratios for \nthp\, are in their optically thin, LTE, values, but for \hcn\ they are not; the $F=1 \to 1$ hyperfine line shows abnormally weak intensities. A simple two-component model shows that self-absorption of the background $F = 1 \to 1$ hyperfine line by the main $F = 2 \to 1$ hyperfine line of a cold, foreground, redshifted cloud can reproduce the observed \hcn\, hyperfine intensities and match the \hcn\, and \hcop\, line profiles. All of these results are consistent with self-absorption of the optically thick lines on the red side of the profile, as expected for collapsing clumps. A simple two-cloud model suggests that this sample represents dense clumps with extreme collapse velocities, $V_{inf} \sim 2.4$ \kms.

Modern studies in astrophysics and cosmology increasingly rely on simulations and cross-survey analyses, yet differences in data generation, instrumentation, calibration, and unmodeled physics introduce distribution mismatches between datasets (``domain shift''). In machine-learning pipelines, this occurs when the joint distribution of inputs and labels differs between the training (source) and application (target) domains, causing source-trained models to underperform on the target. Transfer learning and domain adaptation provide principled ways to mitigate this effect. We study a concrete simulation-to-observation case: semi-supervised domain adaptation (SSDA) to transfer a four-class spectral classifier -- high-redshift quasars, low-redshift quasars, galaxies, and stars -- from J-PAS mock catalogs based on DESI spectra to real J-PAS observations. Our pipeline pretrains on abundant labeled DESI$\rightarrow$J-PAS mocks and adapts to the target domain using a small labeled J-PAS subset. We benchmark SSDA against two baselines: a J-PAS--only supervised model trained with the same target-label budget, and a mocks-only model evaluated on held-out J-PAS data. On this held-out J-PAS data, SSDA achieves a macro-F1 score (balancing precision and recall) of $0.82$ and an overall true positive rate of $0.89$, compared to $0.79/0.85$ for the J-PAS--only baseline and $0.73/0.87$ for the mocks-only model. The gains are driven primarily by improved quasar classification, especially in the high-redshift subclass ($\mathrm{F1}=0.66$ vs.\ $0.55/0.37$), yielding better-calibrated candidate lists for spectroscopic targeting (e.g., WEAVE-QSO) and AGN searches. This study shows how modest target supervision enables robust, data-efficient simulation-to-observation transfer when simulations are plentiful but target labels are scarce.

Massive stars dominate the radiative and mechanical feedback of young stellar populations, yet their intense ultraviolet fields and strong winds are typically presumed to preclude Earth-like habitability. We quantify this expectation by mapping time dependent habitable zones (HZs) for solar-metallicity stars with initial masses of $0.8$-$120\,M_\odot$. From rotating and non-rotating \textsc{GENEC} tracks we derive bolometric ``climate'' HZ boundaries and enforce XUV energy-limited escape and wind ram-pressure retention constraints for a dipole-magnetized Earth analogue. The operational inner edge is set by the most restrictive limit, and we measure the annulus lifetime, the longest continuous residence at fixed orbit, and the maximum number of dynamically packed terrestrial planets it can host. We find a sharp main-sequence ceiling: while a $9\,M_\odot$ star sustains an operational HZ for $\sim 30$~Myr at $\sim 70$-$130$~AU, the main-sequence annulus becomes brief and extremely narrow by $12\,M_\odot$ and disappears by $15\,M_\odot$. Post main-sequence evolution can reopen HZs up to $\sim 25$-$30\,M_\odot$, but only for $\sim 0.03$-$1.5$~Myr at hundreds to $\sim 10^3$~AU, disappearing by $\sim 40\,M_\odot$. Rotation modestly increases habitable lifetimes near the upper main sequence without altering the high mass ceiling. Initial Mass Function (IMF) weighting shows that massive stars contribute only $\sim 10^{-4}$ of the habitable planet-time budget. Even so, they still add of order a few $10^{5}$ operationally habitable Earth analogues to the Milky Way at any instant. This implies that massive star systems are unlikely to dominate the Galaxy wide habitability budget, but they may still provide a set of short-lived, observationally distinct targets for biosignature searches.

The Milky Way hosts astrophysical accelerators capable of producing high-energy cosmic rays. These cosmic rays can interact with the interstellar medium (ISM) across the Galaxy to produce neutrinos and gamma rays (propagation component), while their interactions with ambient material at their acceleration sites, such as supernova remnants, can give rise to the source component of the gamma-ray and neutrino flux. In this paper, we estimate the source component of the Galactic neutrino flux using simulated populations of Galactic gamma-ray sources. We compare our results with observations from neutrino experiments in the energy range of 1-30 TeV. Using simulated populations of Galactic TeV gamma-ray sources, we exploit the correlation between gamma rays and neutrinos and introduce a bracketing approach to constrain the range for the source contribution of the Galactic neutrino flux. For the upper limit, we used a simulation describing the entity of Galactic gamma-ray sources, whereas the lower limit was estimated using the hadronic component of the Galactic supernova remnant population. Our results show that the difference between this maximum and minimum is less than an order of magnitude and the flux range is comparable to the Galactic neutrino flux from the cosmic-ray interaction with the ISM. The results agree with the observed signals from IceCube and ANTARES and suggest that the propagation component, combined with the minimum source contribution predicted by the supernova-remnant model, approaches the observed neutrino flux, leaving little room for significant enhancements of the emission originating from propagating cosmic rays.

Water ice has been detected in several supernova remnants despite the strong heating and radiation in these environments. This challenges standard expectations for dust survival. Using two dimensional magnetohydrodynamic simulations, we study how a supernova shock interacts with a dense interstellar cloud. The simulations show that the shock naturally compresses the cloud into dense structures similar to those inferred in well known remnants. Although temperatures remain high in the adiabatic phase, simple considerations indicate that cooling would act quickly once included. These results suggest that shock cloud interactions create the physical conditions needed for water ice to form. Future work including radiative cooling and grain surface chemistry will allow direct modelling of ice growth in these compressed regions.

The presence of dust in supernova remnants (SNRs) is confirmed by extensive infrared data from observatories such as Spitzer, Herschel, and JWST, alongside theoretical models of dust formation. This study explores the existence of dust and ices, particularly water ice via 62 μm in SNRs such as the Crab Nebula and N49, using observational data and preliminary modeling with Cloudy. Observations suggest that water ice may be present in IC 443 and possibly other remnants, though the 63 μm band could also indicate [OI] emission. Theoretical models indicate that water ice could survive under certain conditions in SNRs, with densities and temperatures analyzed. Further observations and refined simulations are needed to confirm these findings.

GPM 1839-10 is an intriguing long-period radio transient (LPT), distinguished by its activity spanning at least three decades and its highly unusual emission characteristics. These features include orthogonal polarization mode (OPM) switches, down-drifting sub-structures, and distinct linear-to-circular polarization conversion behaviors. In this work, we present follow-up observations utilizing the FAST telescope at L-band, yielding a total of seven detected radio pulses. We find a consistent association between OPM switches and a decrease in polarized intensity. This feature strongly supports the hypothesis that the OPM switches are generated by the incoherent summation of OPMs. Our measured Rotation Measures (RMs) are consistent with previous observations, indicating that the magneto-ionic environment is stable. If the source is in a binary system, such stability suggests it may host a weakly magnetized companion. Crucially, we firstly observe clear evidence of a cyclotron absorption feature in one radio pulse, a signature rarely observed in radio sources. This feature allows us to infer that the magnetic field strength at the absorption site has a lower limit of tens of Gauss, which is necessary for the phenomenon to occur. This characteristic can be explained in a scenario where GPM 1839-10 possesses a weakly magnetized companion star.

Neutron stars in globular clusters follow complex evolutionary pathways shaped by binary interactions, mass transfer, and dynamical exchanges. Direct N-body simulations such as NBODY6++GPU successfully model stellar dynamics and compact object formation, but they usually do not track pulsar spin evolution or magnetic field decay explicitly. Building on Papers I and II of this series, we identify this gap and present a case study from an existing simulation with N = 105000 particles, showing how a neutron star forms and evolves for 200 Myr without any pulsar-physics tracking. We compare this situation with recent implementations and outline a seven-scenario framework that includes magnetic dipole spin-down, exponential magnetic field decay, environmental torques, accretion-driven spin-up, gravitational-wave emission, and merger-driven evolution. As an example, the neutron star we label Pulsar973 forms at t = 800 Myr with a post-supernova mass of 5.35 solar masses and evolves to 2.52 solar masses by t = 1000 Myr, but still lacks period P, period derivative Pdot, magnetic field B, and scenario classification. We provide mathematical formulations and specific integration points within NBODY6++GPU (Hermite scheme, Ahmad-Cohen neighbors, KS regularization, and BSE stellar evolution) to enable scenario-based pulsar evolution within direct N-body simulations.

The observations of protostellar systems show a transition in the radial profile of specific angular momentum (and rotational velocity), evolving from $j\sim{\rm constant}$ ($v_φ\sim r^{-1}$) in the infalling-rotating envelope to $j\propto r^{1/2}$ ($v_φ\sim r^{-1/2}$) in the Keplerian disk. We employ global MHD disk simulations of gravitational collapse starting from a supercritical prestellar core, that forms a disk and envelope structure in a self-consistent manner, in order to determine the physics of the Envelope-Disk Transition Zone (ENDTRANZ). Our numerical results show the transition from the infalling-rotating envelope to Keplerian disk happens through a jump in the $j-r$ profile over a finite radial range, which is characterized by the positive local gravitational torques. The outer edge of the ENDTRANZ is identified where the radial infall speed ($v_r$) begins a sharp decline in magnitude and $j$ begins a transition from $j\sim{\rm constant}$ toward $j\sim r^{1/2}$. Moving radially inward, the centrifugal radius ($r_{\rm CR}$) is defined where $v_φ$ first transitions to Keplerian velocity at the disk's edge. Farther inward of $r_{\rm CR}$, model disk develops a super-Keplerian rotation due to self-gravity. The inner edge of the ENDTRANZ is defined at the centrifugal barrier ($r_{\rm CB}$) where $v_r$ drops to negligible values. Inside $r_{\rm CB}$, a net negative gravitational torque drives mass accretion onto the protostar. On observational grounds, we identify a jump in the observed $j-r$ profile in L1527 IRS for the first time using the ALMA eDisk data. Comparison with the numerical radial behavior from our MHD disk simulations suggests the observed $j-r$ jump can be used as a kinematical tracer for the existence of ENDTRANZ. Our results offer insights into the observable imprint of angular momentum redistribution mechanisms during star-disk formation.

We present photometric and spectroscopic observations of luminous red nova (LRN) AT 2025abao, the fourth discovered in M 31. The LRN, associated to the AGB star WNTR23bzdiq, was discovered during the fast rise following the minimum phase. It reached the peak at $g=15.1$ mag ($M_g=-9.5\pm0.1$ mag), and then it settled onto a long-duration plateau in the red bands, lasting 70 days, while it was slowly linearly declining in the blue bands. The object showed similarities at peak with the canonical LRNe V838 Monocerotis, V1309 Scorpii, and with the faint and fast-evolving AT 2019zhd, the third LRN in M31, though the later evolution is different. Spectroscopically, AT 2025abao evolved as a canonical LRN: the early spectra present a blue continuum with narrow Balmer lines in emission; at peak, the spectral continuum has cooled to a yellow colour, with a photospheric temperature of 6000 K. Balmer lines have weakened while absorption lines from metals (Fe I, Fe II, Sc II, Ba II, Ti II) have developed, and in particular broad (FWHM$\sim$700 km/s) from the UV Ca II H&K lines. Medium- and high-resolution spectra reveal a counter-P Cygni absorption profile in H$α$. Finally, late time spectra show an orange continuum ($T\sim$4000-5000 K), a return in strength of the Balmer lines and the formation of molecular absorption bands. AT 2025abao is the rare case of a LRN with detailed archival information regarding the progenitor system. For the first time, we obtained the spectral energy distribution in the infrared of the precursor of a LRN, which is consistent with that of an M giant/AGB. We propose that the dichotomy of light curve behaviour in LRNe (two peaks vs. plateau) can be explained by the extent and H-richness of the common envelope.

Cold atomic and molecular gas are commonly observed in the winds of both external galaxies and the Milky Way, yet the survival and origin of these cool phases within hot galactic winds is poorly understood. To help gain insight into these problems, we carry out time-dependent chemical modelling of cool clouds in the Milky Way's nuclear wind, which possess unusual molecularto-atomic hydrogen ratios that are inconsistent with both disc values and predictions from chemical equilibrium models. We confirm that CO and Hi emission comparable to that in the observed nuclear wind clouds cannot be produced by gas in chemical equilibrium, but that such conditions can be produced in a molecule-dominated cloud that has had its atomic envelope rapidly removed and has not yet reached a new chemical equilibrium. Clouds in this state harbour large reservoirs of molecular gas and consequently have anomalously large CO-to-H2 conversion factors, suggesting that the masses of the observed clouds may be significantly larger than suggested by earlier analyses assuming disc-like conversions. These findings provide a new framework for interpreting cold gas in galactic winds, providing strong evidence that cold outflows can originate from the galactic disc molecular clouds that survive acceleration into the wind but lose their diffuse atomic envelopes in the process, and suggesting that the Milky Way's nuclear outflow may be more heavily mass-loaded than previously thought.

We present a systematic study of the internal mass structure of early-type galaxies (ETGs) based on 106 galaxy-scale strong gravitational lenses with background quasars, all having spectroscopic redshifts. From this parent sample, we select 24 systems with high-quality ancillary data for joint analysis of strong lensing geometry and stellar kinematics. A key contribution is the derivation of new single-aperture stellar velocity dispersions for 11 lens galaxies via an iterative spectroscopic fitting procedure that mitigates quasar contamination, providing previously unavailable data. We model the total mass-density profile as a power law, $ρ\propto r^{-γ}$, and parameterise its logarithmic slope as $γ= γ_0 + γ_z \cdot z_l + γ_s \cdot \log \tildeΣ$, where $z_l$ is the lens redshift and $\tildeΣ$ the surface mass density. Within a flat $Λ$CDM framework and using DESI BAO measurements as a prior, we constrain the parameters via Monte Carlo nested sampling to $γ_0 = 1.62^{+0.11}_{-0.12}$, $γ_z = -0.35^{+0.08}_{-0.09}$, and $γ_s = 0.37^{+0.08}_{-0.07}$ ($68\%$ confidence intervals). Our results robustly demonstrate that $γ$ increases with surface mass density ($γ_s > 0$) and decreases with redshift ($γ_z < 0$). This implies that, at fixed redshift, galaxies with denser stellar cores have steeper mass profiles, while at fixed density, profiles become shallower at higher redshifts. By successfully applying the joint lensing--dynamics method to a substantial, independently acquired sample of lensed quasars, this work provides crucial validation of structural trends previously observed in galaxy--galaxy lensing systems, reinforcing the established evolutionary picture for massive ETGs and establishing lensed quasars as a potent probe of galaxy structure.

This is a review article for The Review of Particle Physics 2026 (aka the Particle Data Book), appearing as Chapter 25. It forms a compact review of knowledge of the cosmological parameters near the end of 2025. Topics included are Parametrizing the Universe; Extensions to the standard model; Probes; Bringing observations together; Outlook for the future.

The superorbital period of LMC X-4 is among the most stable known in Roche-lobe overflow, high-mass X-ray binaries. We analyzed 33 years of monitoring data from the Compton Gamma Ray Observatory Burst and Transient Source Experiment (CGRO BATSE), the Rossi X-ray Timing Explorer All-Sky Monitor (RXTE ASM), the Neil Gehrels Swift Burst Alert Telescope (Swift BAT), the Monitor of All-sky X-ray Image Gas Slit Camera (MAXI GSC), and the Fermi Gamma-ray Burst Monitor (Fermi GBM). The measured phases show a smooth long-term trend with superposed systematic fluctuations. Fits with cubic, quartic, and sinusoidal models indicate that the quartic and sinusoidal forms provide significantly better descriptions, with the sinusoidal model yielding an $8900^{+210}_{-230}$-day modulation. Such a long timescale is unlikely to arise from orbital motion around a tertiary companion. The fluctuations resemble stochastic, glitch-like events on several-hundred-day timescales. Their rms period variation exceeds that of the smooth trend, yet the total rms period variation over 33 years remains only 0.55\%, demonstrating the exceptional stability of the superorbital period. During MJD 57000-60461, we detect a phase offset of 0.044$\pm$0.010 cycles between the soft and hard X-ray bands. This offset can be reproduced by including a higher-harmonic term in the azimuthal disk model, allowing a transition from antisymmetric to asymmetric structure. A contemporaneous decline in the hard X-ray flux suggests a partial obscuration of the emission region, similar to the anomalous low state in Her X-1. This evolving-disk scenario may also explain the superorbital phase shift previously reported in Her X-1.

With the forth observing run of the LIGO-Virgo-KAGRA gravitational-wave network, which enabled the discovery of the kilonova (KN) counterpart to GW170817, ending with no new confirmed neutron star mergers, the intrinsic rate of these events must be even lower than previously estimated. As a result, building a sample of KNe will remain challenging even with continued GW observations, motivating complementary discovery strategies that do not rely on gravitational-wave triggers. In this work, we consider how leveraging bright short gamma-ray burst afterglows can aid in the discovery on KNe with the Rubin Observatory's upcoming Legacy Survey of Space and Time (LSST), whose unprecedented depth will make such detections feasible. We find that nearly on-axis ($θ_{\rm view} \leq 30°$) afterglows can enhance KN detection rates in the LSST $g$-band from $29^{+51}_{-21} \ \rm yr^{-1}$ to $91^{+160}_{-65} \ \rm yr^{-1}$. We further show how the colors of the observed events can be used to distinguish between neutron star merger counterparts with and without KN emission. This study demonstrates how critical multi-wavelength and multi-survey observations are for these rare events, especially without context from gravitational waves. Fortunately, detectable events will likely be discovered near peak with LSST, allowing for rapid follow-up and confirmation. We discuss key uncertainties in our study, particularly volume rate of merger events, and the degeneracy between the empirically determined explosion energy and ambient medium density.

In this eighth paper of the SOFIA-FORCAST series on Milky Way GHII regions, we present an analysis of the massive star-forming complex W43 Main. We compared our 11 - 37 micron maps with multi-wavelength observations from the near-infrared to radio, and investigated the physical nature of compact sources and dust substructures. We applied SED fitting to constrain properties of the compact infrared objects, and examined the evolutionary states of the extended subregions. We identified 20 compact infrared objects, 16 (80%) of which we classify as massive young stellar objects (MYSOs) or candidate MYSOs. W43 Main resides at the junction of the Scutum spiral arm and the Galactic Bar, a location where enhanced turbulence is anticipated and has been proposed as a potential influence on star-formation activity. Nevertheless, our analysis shows that its Lyman continuum photon production rate, the mass of its most massive MYSO, and its MYSO density are all consistent with the survey-wide median values. We therefore conclude that, despite W43 Main's unique Galactic environment, its present star formation activity appears broadly consistent with that of an average Galactic GHII region.

Measurements from the Absolute Radiometer for Cosmology, Astrophysics, and Diffuse Emission 2 (ARCADE-2) reveal a strong radio background in the GHz frequency range. Since the cosmological 21-cm signal is measured relative to the background radiation temperature, the presence of a radio excess can significantly alter its characteristics. Previous studies have explored the impact of an inhomogeneous radio background on the global 21-cm signal and 21-cm power spectrum. This non-uniform radio background is also expected to introduce substantial non-Gaussianity. In this work, using the bispectrum, we analyze the non-Gaussianity in the 21-cm signal in the presence of an excess galactic radio background and investigate how line-of-sight radio fluctuations from early galaxies influence its nature. We find that even a moderate enhancement in radio efficiency in early galaxies significantly affects the small-scale 21-cm bispectrum. Furthermore, the delayed heating transition caused by a galactic radio background shifts the sign change in the squeezed-limit bispectrum to lower redshifts ($z\sim11$), providing a potential observational signature for distinguishing different radio background models. These results demonstrate that the 21-cm bispectrum, particularly in the squeezed limit, is highly sensitive to radio background fluctuations, making it a powerful tool for probing high-redshift radio-loud sources and the physics of the early cosmic epoch.

The Large High Altitude Air Shower Observatory (LHAASO) monitors sources within its field of view for up to 7 hours daily, achieving a duty cycle exceeding 98% and an annual point-source sensitivity of 1.5% Crab Units (CU) in the very high energy (VHE) band. This unbiased sky-survey mode facilitates systematic monitoring and investigation of outburst phenomena. In this paper, we present results from an unprecedented three-year monitoring campaign (March 2021--March 2024) of Mrk421 using LHAASO, spanning energies from 0.4 TeV to 20 TeV. We find that the blazar stayed in a quiescent state in 2021 and became active starting in 2022 with a total of 23 VHE outburst events identified, where the highest observed daily significance reaches $20\,σ$ with a flux equivalent to approximately 3.3~CU. LHAASO's continuous monitoring suggests the flaring occupancy of Mrk~421 to be around 14%. During long-term monitoring, multiwavelength (MWL) variability and correlation analyses are conducted using complementary data from Fermi-LAT, MAXI-GSC, Swift-XRT, and ZTF. A significant correlation ($>3\,σ$) is observed between X-ray and VHE bands with no detectable time lag, while the correlation between GeV and TeV bands is weaker. The flux distribution of the TeV emission during the quiescent state is different from that in the active state, implying the existence of two modes of energy dissipation in the blazar jet. Using simultaneous MWL data, we also analyzed both the long-term and outburst-period SEDs, and discussed the possible origin of the outburst events.

We examine the multi-wavelength period-luminosity-color relations for Cepheid variables in the Large and Small magellanic Clouds and the Milky Way. From first-principles stellar physics, the luminosity of a Cepheid is determined by its radius and surface temperature, yielding a fundamental PLC relation whose observational proxies are pulsation period and intrinsic color. Using Cepheids in the Magellanic Clouds, we show that the PLC relation recovers the known geometries and line-of-sight tilts of their disks, confirming its ability to detect true distance-modulus variations that are achromatic and consistent across all filters. Surprisingly, for Milky Way Cepheids with individually determined reddenings and HST and Gaia parallaxes, the residuals from multi-wavelength PL fits are also found to be achromatic, identical in sign and amplitude across all passbands, in this case indicating that parallax errors are the dominant source of scatter. Applying bandpass-averaged corrections to individual Cepheids recovers the theoretically expected wavelength-dependent narrowing of the instability strip, and results in revised parallaxes with a median improvement in precision of roughly a factor of two. In addition, they show no statistically significant correlation with metallicity over the range -0.2 < Fe/H < 0.05 dex. The final extinction- and reddening-corrected PLC relation yields an rms scatter of 0.04 mag, corresponding to 2 percent precision in distance per star. Use of a physically grounded PLC will provide a more robust foundation for the Cepheid-based extragalactic distance scale and the determination of the Hubble constant.

At low redshifts, hydroxyl megamasers (OHMs) have been shown to trace galaxy mergers, obscured starbursts, high molecular gas densities, and candidate dual supermassive black hole systems. Given this astrophysical utility, exploring these sources at larger cosmological look-back times is therefore of key interest. While previous OHM surveys have been limited to redshifts of $z \lesssim 0.25$, the ability to expand the OHM frontier is significantly enhanced with new high-sensitivity radio facilities such as MeerKAT. In this Letter, we report the discovery of an OHM in the gravitational lens system HATLAS J142935.3-002836 at $z = 1.027$, the most distant OHM source yet detected. The spectrum has blended 1667 and 1665 MHz emission and exhibits a highly complex profile, with spectral components ranging in widths of $<8$ km s$^{-1}$ to $\sim300$ km s$^{-1}$. The integrated (magnification uncorrected) luminosity of log($L_{\rm OH} / L_{\odot}$) = 5.51 $\pm$ 0.67 makes this the most apparently luminous OHM known to date. In the same wide-band dataset, we have also detected a previously unknown ${\rm H I}$ absorption line. The signal-to-noise ratio of over 150 with just a 4.7 h observation highlights the potential that MeerKAT and the future Square Kilometre Array mid-frequency array offer to explore the high-redshift OHM universe.

Galaxy evolution is driven by star formation and stellar feedback on scales unresolved by current high-resolution cosmological simulations, requiring robust subgrid models. However, these models remain degenerate, often calibrated primarily to match observed stellar masses. To explore these degeneracies, we conduct three state-of-the-art cosmological zoom-in simulations of the same galaxy, each incorporating different subgrid models: mechanical feedback, a combination of mechanical and thermal feedback, and delayed cooling. We compare their circumgalactic media (CGM) through quasar absorption sightlines of HI, MgII, CIV, and OVI. Our findings demonstrate that despite producing galaxies with the same stellar masses, the models lead to distinct feedback modes and CGM properties. Column densities and covering fractions serve as effective diagnostics of subgrid models, with all four ions providing strong constraints as they trace diverse gas phases, exhibit complementary spatial distributions, and originate from different mechanisms. Although all simulations bracket observed column density distributions, direct comparisons are limited by scarce detections and significant scatter in absorption strengths. Covering fractions of weak absorbers provides the most robust constraints. All models fail to reproduce HI and MgII covering fractions, and delayed cooling overproduces OVI covering fractions, while the other models underproduce them. The simulation including mechanical feedback reproduces the observed CIV covering fractions well, whereas the other models show slight offsets. We argue that this discrepancy is likely driven by unresolved thermal structures for HI and MgII, and insufficient metals for CIV and OVI, arising from missing physics such as AGNs or cosmic rays.

Galaxy evolution is shaped by star formation and stellar feedback at scales unresolved by current high-resolution cosmological simulations. Precise subgrid models are thus necessary, and different approaches have been developed. However, they are degenerate and often primarily calibrated to reproduce stellar masses from observations. To explore these degeneracies, we perform three cosmological zoom-in radiation-hydrodynamics simulations of the same galaxy within a $5\times10^{11}\rm\ M_\odot$ dark matter halo at $z\sim1$, each with a different subgrid model: mechanical feedback, a combination of mechanical feedback and thermal feedback, and delayed cooling. We calibrate the simulations to match in stellar mass, isolating the effect of the models on the circumgalactic medium (CGM). Our findings demonstrate that despite producing galaxies with comparable stellar masses, the three models lead to distinct feedback modes, resulting in notable variations in the CGM properties. The delayed cooling run is dominated by ejective feedback and exhibits high burstiness, whereas mechanical and the hybrid models primarily feature preventive feedback, respectively acting at the galaxy and halo scales. Delayed cooling reduces the baryon mass to half the universal baryon fraction while mechanical feedback retains most baryons, with the hybrid model standing in between. Delayed cooling also ejects significantly more metals into the CGM than both other models. While for delayed cooling and mechanical feedback metals are almost evenly distributed in the CGM, they are concentrated around satellites in the hybrid model. These discrepancies emphasize the need to design an appropriate subgrid model to understand how stellar feedback regulates galaxy growth.

The landmark multi-messenger observations of the binary neutron star (BNS) merger GW170817 provided firm evidence that such mergers can produce short gamma-ray bursts (sGRBs). However, the limited number of BNS detections by current gravitational-wave (GW) observatories raises the question of whether BNS mergers alone can account for the full observed sGRB population. We analyze a comprehensive set of 64 BNS population synthesis models with a Monte Carlo-based framework to reproduce the properties of sGRBs detected by Fermi-GBM over the past 16 years. We consider three jet geometry scenarios: a universal structured jet calibrated to GW170817, a universal top-hat jet, and a non-universal top-hat jet with distributions of core opening angles. Our results show that models characterized by low local BNS merger rates ($R_{BNS}(0) \lesssim 50$ Gpc$^{-3}$ yr$^{-1}$) predict too few observable sGRBs to reproduce the Fermi-GBM population, effectively disfavoring them as sole progenitors. Even when relaxing assumptions on jet geometry, low-rate models remain viable only for wide jets ($θ_c \ge 15^\circ$), in tension with the narrow jet cores ($θ_c \approx 6^\circ$) inferred from sGRB afterglow observations. In contrast, models with local merger rates of order $R_{BNS}(0) \approx 100$ Gpc$^{-3}$ yr$^{-1}$ successfully reproduce the observed sGRB population, assuming a plausible fraction of BNS mergers launch relativistic jets and realistic jet geometries. This analysis highlights the power of combining GW observations of BNS mergers with electromagnetic observations of sGRBs to place robust constraints on the BNS merger population and to assess their role as progenitors of sGRBs.

The Lyman $α$ (Ly$α$) line from high-redshift galaxies is a powerful probe of the Epoch of Reionization (EoR). Neutral hydrogen in the intergalactic medium (IGM) can significantly attenuate the emergent Ly$α$ line, even in the damping wing of the cross-section. However, interpreting this damping wing imprint relies on our prior knowledge of the spectrum that escapes from the galaxy and its environs into the IGM. This emergent spectrum is highly sensitive to the composition and geometry of the interstellar and circumgalactic media, and so exhibits a large galaxy to galaxy scatter. Characterizing this scatter is further complicated by non-trivial selection effects introduced by observational surveys. Here we build a flexible, empirical model for the emergent Ly$α$ spectra. Our model characterizes the emergent Ly$α$ luminosity, the velocity offset of the Ly$α$ line with respect to the systemic redshift, and the H$α$ luminosity, with multivariate probability distributions conditioned on the UV magnitude. We constrain these distributions using $z\sim5-6$ galaxy observations with VLT MUSE and JWST NIRCam, forward-modeling observational selection functions together with galaxy parameters. Our model results in Ly$α$ equivalent width distributions that are a better match to (independent) Subaru observations than previous empirical models. The extended distributions of Ly$α$ equivalent widths and velocity offsets we obtain could facilitate Ly$α$ transmission during the early stages of the EoR. We also illustrate how our model can be used to identify GN-z11-like outliers, potentially originating from merging systems. We publish fitting functions and make our model publicly available.

Globular cluster stellar streams probe galaxy-formation processes and can potentially reveal the distribution of dark matter in galaxies. In many theoretical studies, streams are modeled with particle-spray or direct N-body codes. But particle-spray methods abstract away the internal dynamics of the progenitor by making strong assumptions about the escape physics, while direct N-body is prohibitively expensive for realistic (N>10^5) systems. In this paper, we present the stream-modeling capabilities of KRIOS, a new basis-expansion N-body code for collisional stellar dynamics, that bridges this runtime vs. accuracy gap. We show that KRIOS reproduces NBODY6++GPU cluster models, and their associated streams, more accurately than particle spray in a fraction of the NBODY6++GPU wall-clock time. We then compare KRIOS to various particle-spray methods on 10 orbits similar to known Milky Way streams. The morphology and kinematics of these streams most disagree when the progenitor is tightly bound to the host, as these systems are often subject to stronger tidal forces. Finally, we discuss which elements of the progenitor physics are most important for modeling stellar streams and how these might be incorporated into particle-spray methods.

I consider recent observations of ultramassive black holes. These appear to confirm theoretical predictions that the relation between central black hole mass $M$ and spheroid velocity dispersion $σ$ has the same form $M \propto σ^4$ in spiral galaxies, elliptical galaxies, and cluster ellipticals, but has differing normalizations. These arise from the need for longer black hole accretion episodes to expel the gas otherwise potentially able to feed the holes in the latter two types of host. In a sample drawn from a mixture of galaxy host types the fitted power of $σ$ will slightly exceed the theoretically-derived value of 4 because of the differing normalizations. The observed hole masses do not currently reach the theoretical maximum values possible for disc accretion, set by the equality of the ISCO and self-gravity radii, probably because the host galaxies have insufficient gas.

We investigate the velocity evolution of free-floating planets and interstellar objects (``free floaters'') through gravitational scatterings by field stars (with the stellar mass $m$ much larger than the mass of the floater, $m_p$). We show that the equilibrium velocity -- where dynamical friction balances stochastic acceleration -- is given by $σ\sqrt{2\ln(m/m_p)}$ (where $σ$ is the velocity disperson of the field stars), diverging from the standard energy equipartition scaling. While the timescale to reach this equilibrium is prohibitively long, we find that slow floaters ($v \lesssim σ$) undergo mass-independent acceleration, doubling their velocities within a few relaxation times. Consequently, free floaters initially following the Maxwellian distribution of their parent stars develop distinctly non-Maxwellian velocity distributions on a relaxation timescale. Since the relaxation time of the Galactic disk is longer than the age, our results suggest that the kinematics of low-mass free floaters in the disk may preserve signatures of their parent stars and ejection history.