arXiv Daily Digest - 2026-02-03

Large language models often exhibit increased sycophantic behavior after preference-based post-training, showing a stronger tendency to affirm a user's stated or implied belief even when this conflicts with factual accuracy or sound judgment. We present a formal analysis of how alignment from human feedback can increase this failure mode by identifying an explicit amplification mechanism that causally links optimization against a learned reward to bias in the human preference data used for alignment. We show that the direction of behavioral drift is determined by a covariance under the base policy between endorsing the belief signal in the prompt and the learned reward, and that the first-order effect reduces to a simple mean-gap condition. We then analyze reward learning from pairwise comparisons under random utility models like Bradley-Terry and characterize when bias in human annotators' preferences induces this reward gap. Next, we propose a training-time intervention designed to neutralize the amplification mechanism itself. Among all post-trained policies that prevent sycophantic behavior from increasing, we characterize the unique policy closest in KL divergence to the unconstrained post-trained policy, and derive the corresponding minimal reward correction as a closed-form agreement penalty. Computational experiments find that reward gaps are common and cause behavioral drift in all the configurations considered.

Ultrasound imaging is the primary diagnostic modality for detecting Gallbladder diseases due to its non-invasive nature, affordability, and wide accessibility. However, the low resolution and speckle noise inherent to ultrasound images hinder diagnostic reliability, prompting the use of large convolutional neural networks that are difficult to deploy in routine clinical settings. In this work, we propose CortiNet, a lightweight, cortical-inspired dual-stream neural architecture for gallbladder disease diagnosis that integrates physically interpretable multi-scale signal decomposition with perception-driven feature learning. Inspired by parallel processing pathways in the human visual cortex, CortiNet explicitly separates low-frequency structural information from high-frequency perceptual details and processes them through specialized encoding streams. By operating directly on structured, frequency-selective representations rather than raw pixel intensities, the architecture embeds strong physics-based inductive bias, enabling efficient feature learning with a significantly reduced parameter footprint. A late-stage cortical-style fusion mechanism integrates complementary structural and textural cues while preserving computational efficiency. Additionally, we propose a structure-aware explainability framework wherein gradient-weighted class activation mapping is only applied to the structural branch of the proposed CortiNet architecture. This choice allows the model to only focus on the structural features, making it robust against speckle noise. We evaluate CortiNet on 10,692 expert-annotated images spanning nine clinically relevant gallbladder disease categories. Experimental results demonstrate that CortiNet achieves high diagnostic accuracy (98.74%) with only a fraction of the parameters required by conventional deep convolutional models.

Recent large language models (LLMs) perform strongly on mathematical benchmarks yet often misapply lemmas, importing conclusions without validating assumptions. We formalize lemma$-$judging as a structured prediction task: given a statement and a candidate lemma, the model must output a precondition check and a conclusion$-$utility check, from which a usefulness decision is derived. We present RULES, which encodes this specification via a two$-$section output and trains with reinforcement learning plus section$-$aware loss masking to assign penalty to the section responsible for errors. Training and evaluation draw on diverse natural language and formal proof corpora; robustness is assessed with a held$-$out perturbation suite; and end$-$to$-$end evaluation spans competition$-$style, perturbation$-$aligned, and theorem$-$based problems across various LLMs. Results show consistent in$-$domain gains over both a vanilla model and a single$-$label RL baseline, larger improvements on applicability$-$breaking perturbations, and parity or modest gains on end$-$to$-$end tasks; ablations indicate that the two$-$section outputs and section$-$aware reinforcement are both necessary for robustness.

Automatic Prompt Optimization (APO) is a powerful approach for extracting performance from large language models without modifying their weights. Many existing methods rely on trial-and-error, testing different prompts or in-context examples until a good configuration emerges, often consuming substantial compute. Recently, natural language feedback derived from execution logs has shown promise as a way to identify how prompts can be improved. However, most prior approaches operate in a bottom-up manner, iteratively adjusting the prompt based on feedback from individual problems, which can cause them to lose the global perspective. In this work, we propose Error Taxonomy-Guided Prompt Optimization (ETGPO), a prompt optimization algorithm that adopts a top-down approach. ETGPO focuses on the global failure landscape by collecting model errors, categorizing them into a taxonomy, and augmenting the prompt with guidance targeting the most frequent failure modes. Across multiple benchmarks spanning mathematics, question answering, and logical reasoning, ETGPO achieves accuracy that is comparable to or better than state-of-the-art methods, while requiring roughly one third of the optimization-phase token usage and evaluation budget.

Complex question answering across text, tables and images requires integrating diverse information sources. A framework supporting specialized processing with coordination and interpretability is needed. We introduce DeALOG, a decentralized multi-agent framework for multimodal question answering. It uses specialized agents: Table, Context, Visual, Summarizing and Verification, that communicate through a shared natural-language log as persistent memory. This log-based approach enables collaborative error detection and verification without central control, improving robustness. Evaluations on FinQA, TAT-QA, CRT-QA, WikiTableQuestions, FeTaQA, and MultiModalQA show competitive performance. Analysis confirms the importance of the shared log, agent specialization, and verification for accuracy. DeALOG, provides a scalable approach through modular components using natural-language communication.

Agentic Reinforcement Learning (ARL) focuses on training large language models (LLMs) to interleave reasoning with external tool execution to solve complex tasks. Most existing ARL methods train a single shared model parameters to support both reasoning and tool use behaviors, implicitly assuming that joint training leads to improved overall agent performance. Despite its widespread adoption, this assumption has rarely been examined empirically. In this paper, we systematically investigate this assumption by introducing a Linear Effect Attribution System(LEAS), which provides quantitative evidence of interference between reasoning and tool-use behaviors. Through an in-depth analysis, we show that these two capabilities often induce misaligned gradient directions, leading to training interference that undermines the effectiveness of joint optimization and challenges the prevailing ARL paradigm. To address this issue, we propose Disentangled Action Reasoning Tuning(DART), a simple and efficient framework that explicitly decouples parameter updates for reasoning and tool-use via separate low-rank adaptation modules. Experimental results show that DART consistently outperforms baseline methods with averaged 6.35 percent improvements and achieves performance comparable to multi-agent systems that explicitly separate tool-use and reasoning using a single model.

End-to-end autonomous driving models increasingly benefit from large vision--language models for semantic understanding, yet ensuring safe and accurate operation under long-tail conditions remains challenging. These challenges are particularly prominent in long-tail mixed-traffic scenarios, where autonomous vehicles must interact with heterogeneous road users, including human-driven vehicles and vulnerable road users, under complex and uncertain conditions. This paper proposes HERMES, a holistic risk-aware end-to-end multimodal driving framework designed to inject explicit long-tail risk cues into trajectory planning. HERMES employs a foundation-model-assisted annotation pipeline to produce structured Long-Tail Scene Context and Long-Tail Planning Context, capturing hazard-centric cues together with maneuver intent and safety preference, and uses these signals to guide end-to-end planning. HERMES further introduces a Tri-Modal Driving Module that fuses multi-view perception, historical motion cues, and semantic guidance, ensuring risk-aware accurate trajectory planning under long-tail scenarios. Experiments on the real-world long-tail dataset demonstrate that HERMES consistently outperforms representative end-to-end and VLM-driven baselines under long-tail mixed-traffic scenarios. Ablation studies verify the complementary contributions of key components.

Prediction sets can wrap around any ML model to cover unknown test outcomes with a guaranteed probability. Yet, it remains unclear how to use them optimally for downstream decision-making. Here, we propose a decision-theoretic framework that seeks to minimize the expected loss (risk) against a worst-case distribution consistent with the prediction set's coverage guarantee. We first characterize the minimax optimal policy for a fixed prediction set, showing that it balances the worst-case loss inside the set with a penalty for potential losses outside the set. Building on this, we derive the optimal prediction set construction that minimizes the resulting robust risk subject to a coverage constraint. Finally, we introduce Risk-Optimal Conformal Prediction (ROCP), a practical algorithm that targets these risk-minimizing sets while maintaining finite-sample distribution-free marginal coverage. Empirical evaluations on medical diagnosis and safety-critical decision-making tasks demonstrate that ROCP reduces critical mistakes compared to baselines, particularly when out-of-set errors are costly.

Modeling non-stationary processes, where statistical properties vary across the input domain, is a critical challenge in machine learning; yet most scalable methods rely on a simplifying assumption of stationarity. This forces a difficult trade-off: use expressive but computationally demanding models like Deep Gaussian Processes, or scalable but limited methods like Random Fourier Features (RFF). We close this gap by introducing Random Wavelet Features (RWF), a framework that constructs scalable, non-stationary kernel approximations by sampling from wavelet families. By harnessing the inherent localization and multi-resolution structure of wavelets, RWF generates an explicit feature map that captures complex, input-dependent patterns. Our framework provides a principled way to generalize RFF to the non-stationary setting and comes with a comprehensive theoretical analysis, including positive definiteness, unbiasedness, and uniform convergence guarantees. We demonstrate empirically on a range of challenging synthetic and real-world datasets that RWF outperforms stationary random features and offers a compelling accuracy-efficiency trade-off against more complex models, unlocking scalable and expressive kernel methods for a broad class of real-world non-stationary problems.

In this paper, we identify a sparse reward subsystem within the hidden states of Large Language Models (LLMs), drawing an analogy to the biological reward subsystem in the human brain. We demonstrate that this subsystem contains value neurons that represent the model's internal expectation of state value, and through intervention experiments, we establish the importance of these neurons for reasoning. Our experiments reveal that these value neurons are robust across diverse datasets, model scales, and architectures; furthermore, they exhibit significant transferability across different datasets and models fine-tuned from the same base model. By examining cases where value predictions and actual rewards diverge, we identify dopamine neurons within the reward subsystem which encode reward prediction errors (RPE). These neurons exhibit high activation when the reward is higher than expected and low activation when the reward is lower than expected.

Reinforcement learning with verifiable rewards has emerged as a promising paradigm for enhancing the reasoning capabilities of large language models particularly in mathematics. Current approaches in this domain present a clear trade-off: PPO-style methods (e.g., GRPO/DAPO) offer training stability but exhibit slow learning trajectories due to their trust-region constraints on policy updates, while REINFORCE-style approaches (e.g., CISPO) demonstrate improved learning efficiency but suffer from performance instability as they clip importance sampling weights while still permitting non-zero gradients outside the trust-region. To address these limitations, we introduce DISPO, a simple yet effective REINFORCE-style algorithm that decouples the up-clipping and down-clipping of importance sampling weights for correct and incorrect responses, yielding four controllable policy update regimes. Through targeted ablations, we uncover how each regime impacts training: for correct responses, weights >1 increase the average token entropy (i.e., exploration) while weights <1 decrease it (i.e., distillation) -- both beneficial but causing gradual performance degradation when excessive. For incorrect responses, overly restrictive clipping triggers sudden performance collapse through repetitive outputs (when weights >1) or vanishing response lengths (when weights <1). By separately tuning these four clipping parameters, DISPO maintains the exploration-distillation balance while preventing catastrophic failures, achieving 61.04% on AIME'24 (vs. 55.42% CISPO and 50.21% DAPO) with similar gains across various benchmarks and models.

Visual robustness and neural alignment remain critical challenges in developing artificial agents that can match biological vision systems. We present the winning approaches from Team HCMUS_TheFangs for both tracks of the NeurIPS 2025 Mouse vs. AI: Robust Visual Foraging Competition. For Track 1 (Visual Robustness), we demonstrate that architectural simplicity combined with targeted components yields superior generalization, achieving 95.4% final score with a lightweight two-layer CNN enhanced by Gated Linear Units and observation normalization. For Track 2 (Neural Alignment), we develop a deep ResNet-like architecture with 16 convolutional layers and GLU-based gating that achieves top-1 neural prediction performance with 17.8 million parameters. Our systematic analysis of ten model checkpoints trained between 60K to 1.14M steps reveals that training duration exhibits a non-monotonic relationship with performance, with optimal results achieved around 200K steps. Through comprehensive ablation studies and failure case analysis, we provide insights into why simpler architectures excel at visual robustness while deeper models with increased capacity achieve better neural alignment. Our results challenge conventional assumptions about model complexity in visuomotor learning and offer practical guidance for developing robust, biologically-inspired visual agents.

Spoken question-answering (SQA) systems relying on automatic speech recognition (ASR) often struggle with accurately recognizing medical terminology. To this end, we propose MedSpeak, a novel knowledge graph-aided ASR error correction framework that refines noisy transcripts and improves downstream answer prediction by leveraging both semantic relationships and phonetic information encoded in a medical knowledge graph, together with the reasoning power of LLMs. Comprehensive experimental results on benchmarks demonstrate that MedSpeak significantly improves the accuracy of medical term recognition and overall medical SQA performance, establishing MedSpeak as a state-of-the-art solution for medical SQA. The code is available at https://github.com/RainieLLM/MedSpeak.

Large language models (LLMs) have demonstrated remarkable potential for automatic short answer grading (ASAG), significantly boosting student assessment efficiency and scalability in educational scenarios. However, their vulnerability to adversarial manipulation raises critical concerns about automatic grading fairness and reliability. In this paper, we introduce GradingAttack, a fine-grained adversarial attack framework that systematically evaluates the vulnerability of LLM based ASAG models. Specifically, we align general-purpose attack methods with the specific objectives of ASAG by designing token-level and prompt-level strategies that manipulate grading outcomes while maintaining high camouflage. Furthermore, to quantify attack camouflage, we propose a novel evaluation metric that balances attack success and camouflage. Experiments on multiple datasets demonstrate that both attack strategies effectively mislead grading models, with prompt-level attacks achieving higher success rates and token-level attacks exhibiting superior camouflage capability. Our findings underscore the need for robust defenses to ensure fairness and reliability in ASAG. Our code and datasets are available at https://anonymous.4open.science/r/GradingAttack.

Evolutionary success depends on the capacity to adapt: organisms must respond to environmental challenges through both genetic innovation and lifetime learning. The gene-centric paradigm attributes evolutionary causality exclusively to genes, while Denis Noble's phenotype-first framework argues that organisms are active agents capable of interpreting genetic resources, learning from experience, and shaping their own development. However, this framework has remained philosophically intuitive but algorithmically opaque. We show for the first time that organismal agency can be implemented as a concrete computational process through heritable phenotypic patterns. We introduce the Phenopoiesis Algorithm, where organisms inherit not just genes but also successful phenotypic patterns discovered during lifetime learning. Through experiments in changing environments, these pattern-inheriting organisms achieve 3.4 times faster adaptation compared to gene-centric models. Critically, these gains require cross-generational inheritance of learned patterns rather than within-lifetime learning alone. We conclude that organismal agency is not a philosophical abstraction but an algorithmic mechanism with measurable adaptive value. The mechanism works through compositional reuse: organisms discover how to compose primitive elements into solutions, encode those compositional recipes, and transmit them to offspring. Evolution operates across multiple timescales -- fast, reversible phenotypic inheritance and slow, permanent genetic inheritance -- providing adaptive flexibility that single-channel mechanisms cannot achieve.

Large language models (LLMs) are increasingly deployed in domains where errors carry high social, scientific, or safety costs. Yet standard confidence estimators, such as token likelihood, semantic similarity and multi-sample consistency, remain brittle under distribution shift, domain-specialised text, and compute limits. In this work, we present Structural Confidence, a single-pass, model-agnostic framework that enhances output correctness prediction based on multi-scale structural signals derived from a model's final-layer hidden-state trajectory. By combining spectral, local-variation, and global shape descriptors, our method captures internal stability patterns that are missed by probabilities and sentence embeddings. We conduct extensive, cross-domain evaluation across four heterogeneous benchmarks-FEVER (fact verification), SciFact (scientific claims), WikiBio-hallucination (biographical consistency), and TruthfulQA (truthfulness-oriented QA). Our Structural Confidence framework demonstrates strong performance compared with established baselines in terms of AUROC and AUPR. More importantly, unlike sampling-based consistency methods which require multiple stochastic generations and an auxiliary model, our approach uses a single deterministic forward pass, offering a practical basis for efficient, robust post-hoc confidence estimation in socially impactful, resource-constrained LLM applications.

Label-supervised manifold alignment bridges the gap between unsupervised and correspondence-based paradigms by leveraging shared label information to align multimodal datasets. Still, most existing methods rely on Euclidean geometry to model intra-domain relationships. This approach can fail when features are only weakly related to the task of interest, leading to noisy, semantically misleading structure and degraded alignment quality. To address this limitation, we introduce FoSTA (Forest-guided Semantic Transport Alignment), a scalable alignment framework that leverages forest-induced geometry to denoise intra-domain structure and recover task-relevant manifolds prior to alignment. FoSTA builds semantic representations directly from label-informed forest affinities and aligns them via fast, hierarchical semantic transport, capturing meaningful cross-domain relationships. Extensive comparisons with established baselines demonstrate that FoSTA improves correspondence recovery and label transfer on synthetic benchmarks and delivers strong performance in practical single-cell applications, including batch correction and biological conservation.

The distributed nature of microservice architecture introduces significant resilience challenges. Traditional testing methods, limited by extensive manual effort and oversimplified test environments, fail to capture production system complexity. To address these limitations, we present Cast, an automated, end-to-end framework for microservice resilience testing in production. It achieves high test fidelity by replaying production traffic against a comprehensive library of application-level faults to exercise internal error-handling logic. To manage the combinatorial test space, Cast employs a complexity-driven strategy to systematically prune redundant tests and prioritize high-value tests targeting the most critical service execution paths. Cast automates the testing lifecycle through a three-phase pipeline (i.e., startup, fault injection, and recovery) and uses a multi-faceted oracle to automatically verify system resilience against nuanced criteria. Deployed in Huawei Cloud for over eight months, Cast has been adopted by many service teams to proactively address resilience vulnerabilities. Our analysis on four large-scale applications with millions of traces reveals 137 potential vulnerabilities, with 89 confirmed by developers. To further quantify its performance, Cast is evaluated on a benchmark set of 48 reproduced bugs, achieving a high coverage of 90%. The results show that Cast is a practical and effective solution for systematically improving the reliability of industrial microservice systems.

Agents powered by large language models (LLMs) are increasingly deployed in settings where communication shapes high-stakes decisions, making a principled understanding of strategic communication essential. Prior work largely studies either unverifiable cheap-talk or fully verifiable disclosure, failing to capture realistic domains in which information has probabilistic credibility. We introduce MixTalk, a strategic communication game for LLM-to-LLM interaction that models information credibility. In MixTalk, a sender agent strategically combines verifiable and unverifiable claims to communicate private information, while a receiver agent allocates a limited budget to costly verification and infers the underlying state from prior beliefs, claims, and verification outcomes. We evaluate state-of-the-art LLM agents in large-scale tournaments across three realistic deployment settings, revealing their strengths and limitations in reasoning about information credibility and the explicit behavior that shapes these interactions. Finally, we propose Tournament Oracle Policy Distillation (TOPD), an offline method that distills tournament oracle policy from interaction logs and deploys it in-context at inference time. Our results show that TOPD significantly improves receiver robustness to persuasion.

Training Large Language Models (LLMs) at ultra-low precision is critically impeded by instability rooted in the conflict between discrete quantization constraints and the intrinsic heavy-tailed spectral nature of linguistic data. By formalizing the connection between Zipfian statistics and random matrix theory, we prove that the power-law decay in the singular value spectra of embeddings is a fundamental requisite for semantic encoding. We derive theoretical bounds showing that uniform quantization introduces a noise floor that disproportionately truncates this spectral tail, which induces spectral flattening and a strictly provable increase in the stable rank of representations. Empirical validation across diverse architectures including GPT-2 and TinyLlama corroborates that this geometric degradation precipitates representational collapse. This work not only quantifies the spectral sensitivity of LLMs but also establishes spectral fidelity as a necessary condition for stable low-bit optimization.

Multi-agent large language model systems can tackle complex multi-step tasks by decomposing work and coordinating specialized behaviors. However, current coordination mechanisms typically rely on statically assigned roles and centralized controllers. As agent pools and task distributions evolve, these design choices lead to inefficient routing, poor adaptability, and fragile fault recovery capabilities. We introduce Symphony-Coord, a decentralized multi-agent framework that transforms agent selection into an online multi-armed bandit problem, enabling roles to emerge organically through interaction. The framework employs a two-stage dynamic beacon protocol: (i) a lightweight candidate screening mechanism to limit communication and computational overhead; (ii) an adaptive LinUCB selector that routes subtasks based on context features derived from task requirements and agent states, continuously optimized through delayed end-to-end feedback. Under standard linear realizability assumptions, we provide sublinear regret bounds, indicating the system converges toward near-optimal allocation schemes. Validation through simulation experiments and real-world large language model benchmarks demonstrates that Symphony-Coord not only enhances task routing efficiency but also exhibits robust self-healing capabilities in scenarios involving distribution shifts and agent failures, achieving a scalable coordination mechanism without predefined roles.

Scientific machine learning (SciML) increasingly requires models that capture multimodal conditional uncertainty arising from ill-posed inverse problems, multistability, and chaotic dynamics. While recent work has favored highly expressive implicit generative models such as diffusion and flow-based methods, these approaches are often data-hungry, computationally costly, and misaligned with the structured solution spaces frequently found in scientific problems. We demonstrate that Mixture Density Networks (MDNs) provide a principled yet largely overlooked alternative for multimodal uncertainty quantification in SciML. As explicit parametric density estimators, MDNs impose an inductive bias tailored to low-dimensional, multimodal physics, enabling direct global allocation of probability mass across distinct solution branches. This structure delivers strong data efficiency, allowing reliable recovery of separated modes in regimes where scientific data is scarce. We formalize these insights through a unified probabilistic framework contrasting explicit and implicit distribution networks, and demonstrate empirically that MDNs achieve superior generalization, interpretability, and sample efficiency across a range of inverse, multistable, and chaotic scientific regression tasks.

Large Language Models (LLMs) can be seen as compressed knowledge bases, but it remains unclear what knowledge they truly contain and how far their knowledge boundaries extend. Existing benchmarks are mostly static and provide limited support for systematic knowledge probing. In this paper, we propose an interactive agentic framework to systematically extract and quantify the knowledge of LLMs. Our method includes four adaptive exploration policies to probe knowledge at different granularities. To ensure the quality of extracted knowledge, we introduce a three-stage knowledge processing pipeline that combines vector-based filtering to remove exact duplicates, LLM-based adjudication to resolve ambiguous semantic overlaps, and domain-relevance auditing to retain valid knowledge units. Through extensive experiments, we find that recursive taxonomy is the most effective exploration strategy. We also observe a clear knowledge scaling law, where larger models consistently extract more knowledge. In addition, we identify a Pass@1-versus-Pass@k trade-off: domain-specialized models achieve higher initial accuracy but degrade rapidly, while general-purpose models maintain stable performance during extended extraction. Finally, our results show that differences in training data composition lead to distinct and measurable knowledge profiles across model families.

Model adaptation to production environment is critical for reliable Machine Learning Operations (MLOps), less attention is paid to developing systematic framework for updating the ML models when they fail under data drift. This paper compares the transfer learning enabled model update strategies including ensemble transfer learning (ETL), all-layers transfer learning (ALTL), and last-layer transfer learning (LLTL) for updating the failed feedforward artificial neural network (ANN) model. The flue gas differential pressure across the air preheater unit installed in a 660 MW thermal power plant is analyzed as a case study since it mimics the batch processes due to load cycling in the power plant. Updating the failed ANN model by three transfer learning techniques reveals that ETL provides relatively higher predictive accuracy for the batch size of 5 days than those of LLTL and ALTL. However, ALTL is found to be suitable for effective update of the model trained on large batch size (8 days). A mixed trend is observed for computational requirement (hyperparameter tuning and model training) of model update techniques for different batch sizes. These fundamental and empiric insights obtained from the batch process-based industrial case study can assist the MLOps practitioners in adapting the failed models to data drifts for the accurate monitoring of industrial processes.

Early and accurate diagnosis of Alzheimer's disease (AD) remains a critical challenge in neuroimaging-based clinical decision support systems. In this work, we propose a novel hybrid deep learning framework that integrates Topological Data Analysis (TDA) with a DenseNet121 backbone for four-class Alzheimer's disease classification using structural MRI data from the OASIS dataset. TDA is employed to capture complementary topological characteristics of brain structures that are often overlooked by conventional neural networks, while DenseNet121 efficiently learns hierarchical spatial features from MRI slices. The extracted deep and topological features are fused to enhance class separability across the four AD stages. Extensive experiments conducted on the OASIS-1 Kaggle MRI dataset demonstrate that the proposed TDA+DenseNet121 model significantly outperforms existing state-of-the-art approaches. The model achieves an accuracy of 99.93% and an AUC of 100%, surpassing recently published CNN-based, transfer learning, ensemble, and multi-scale architectures. These results confirm the effectiveness of incorporating topological insights into deep learning pipelines and highlight the potential of the proposed framework as a robust and highly accurate tool for automated Alzheimer's disease diagnosis.

Preference optimization methods such as DPO align large language models (LLMs) using paired comparisons, but their effectiveness can be highly sensitive to the quality and difficulty of preference pairs. A common heuristic treats small-margin (ambiguous) pairs as noisy and filters them out. In this paper, we revisit this assumption and show that pair difficulty interacts strongly with the optimization objective: when trained with preference-based losses, difficult pairs can destabilize training and harm alignment, yet these same pairs still contain useful supervision signals when optimized with supervised fine-tuning (SFT). Motivated by this observation, we propose MixDPO, a simple yet effective difficulty-aware training strategy that (i) orders preference data from easy to hard (a curriculum over margin-defined difficulty), and (ii) routes difficult pairs to an SFT objective while applying a preference loss to easy pairs. This hybrid design provides a practical mechanism to leverage ambiguous pairs without incurring the optimization failures often associated with preference losses on low-margin data. Across three LLM-judge benchmarks, MixDPO consistently improves alignment over DPO and a range of widely-used variants, with particularly strong gains on AlpacaEval~2 length-controlled (LC) win rate.

Despite significant progress in computational pathology, many AI models remain black-box and difficult to interpret, posing a major barrier to clinical adoption due to limited transparency and explainability. This has motivated continued interest in engineered image-based biomarkers, which offer greater interpretability but are often proposed based on anecdotal evidence or fragmented prior literature rather than systematic biological validation. We introduce SAGE (Structured Agentic system for hypothesis Generation and Evaluation), an agentic AI system designed to identify interpretable, engineered pathology biomarkers by grounding them in biological evidence. SAGE integrates literature-anchored reasoning with multimodal data analysis to correlate image-derived features with molecular biomarkers, such as gene expression, and clinically relevant outcomes. By coordinating specialized agents for biological contextualization and empirical hypothesis validation, SAGE prioritizes transparent, biologically supported biomarkers and advances the clinical translation of computational pathology.

The trade-off between labeled data availability and downstream accuracy remains a central challenge in fine-tuning large language models (LLMs). We propose a principled framework for \emph{budget-aware supervised fine-tuning} by casting LLM adaptation as a contextual Stackelberg game. In our formulation, the learner (leader) commits to a scoring policy and a label-querying strategy, while an adaptive environment (follower) selects challenging supervised alternatives in response. To explicitly address label efficiency, we incorporate a finite supervision budget directly into the learning objective. Our algorithm operates in the full-feedback regime and achieves $\tilde{O}(d\sqrt{T})$ regret under standard linear contextual assumptions. We extend the framework with a Largest-Latency-First (LLF) confidence gate that selectively queries labels, achieving a budget-aware regret bound of $\tilde{O}(\sqrt{dB} + c\sqrt{B})$ with $B=βT$.

We introduce online Hierarchical Task Network (HTN) agents whose behaviors are governed by a set of built-in directives \D. Like other agents that are capable of rebellion (i.e., {\it intelligent disobedience}), our agents will, under some conditions, not perform a user-assigned task and instead act in ways that do not meet a user's expectations. Our work combines three concepts: HTN planning, online planning, and the directives \D, which must be considered when performing user-assigned tasks. We investigate two agent variants: (1) a Nonadaptive agent that stops execution if it finds itself in violation of \D~ and (2) an Adaptive agent that, in the same situation, instead modifies its HTN plan to search for alternative ways to achieve its given task. We present R-HTN (for: Rebellious-HTN), a general algorithm for online HTN planning under directives \D. We evaluate R-HTN in two task domains where the agent must not violate some directives for safety reasons or as dictated by their personality traits. We found that R-HTN agents never violate directives, and aim to achieve the user-given goals if feasible though not necessarily as the user expected.

Large language models are increasingly used for mental health support, yet their conversational coherence alone does not ensure clinical appropriateness. Existing general-purpose safeguards often fail to distinguish between therapeutic disclosures and genuine clinical crises, leading to safety failures. To address this gap, we introduce a clinically grounded risk taxonomy, developed in collaboration with PhD-level psychologists, that identifies actionable harm (e.g., self-harm and harm to others) while preserving space for safe, non-crisis therapeutic content. We release MindGuard-testset, a dataset of real-world multi-turn conversations annotated at the turn level by clinical experts. Using synthetic dialogues generated via a controlled two-agent setup, we train MindGuard, a family of lightweight safety classifiers (with 4B and 8B parameters). Our classifiers reduce false positives at high-recall operating points and, when paired with clinician language models, help achieve lower attack success and harmful engagement rates in adversarial multi-turn interactions compared to general-purpose safeguards. We release all models and human evaluation data.

Conventional financial strategy evaluation relies on isolated backtests in static environments. Such evaluations assess each policy independently, overlook correlations and interactions, and fail to explain why strategies ultimately persist or vanish in evolving markets. We shift to an ecological perspective, where trading strategies are modeled as adaptive agents that interact and learn within a shared market. Instead of proposing a new strategy, we present FinEvo, an ecological game formalism for studying the evolutionary dynamics of multi-agent financial strategies. At the individual level, heterogeneous ML-based traders-rule-based, deep learning, reinforcement learning, and large language model (LLM) agents-adapt using signals such as historical prices and external news. At the population level, strategy distributions evolve through three designed mechanisms-selection, innovation, and environmental perturbation-capturing the dynamic forces of real markets. Together, these two layers of adaptation link evolutionary game theory with modern learning dynamics, providing a principled environment for studying strategic behavior. Experiments with external shocks and real-world news streams show that FinEvo is both stable for reproducibility and expressive in revealing context-dependent outcomes. Strategies may dominate, collapse, or form coalitions depending on their competitors-patterns invisible to static backtests. By reframing strategy evaluation as an ecological game formalism, FinEvo provides a unified, mechanism-level protocol for analyzing robustness, adaptation, and emergent dynamics in multi-agent financial markets, and may offer a means to explore the potential impact of macroeconomic policies and financial regulations on price evolution and equilibrium.

Chat has become the default interface for AI-assisted data analysis. For multi-step, state-dependent analytical tasks, this is a mistake. Building on Woods (1984) Keyhole Effect, the cognitive cost of viewing large information spaces through narrow viewports, I show that chat interfaces systematically degrade analytical performance through five mechanisms: (1) constant content displacement defeats hippocampal spatial memory systems; (2) hidden state variables exceed working memory capacity (approximately 4 chunks under load); (3) forced verbalization triggers verbal overshadowing, degrading visual pattern recognition; (4) linear text streams block epistemic action and cognitive offloading; (5) serialization penalties scale with data dimensionality. I formalize cognitive overload as O = max(0, m - v - W) where m is task-relevant items, v is visible items, and W is working memory capacity. When O > 0, error probability increases and analytical biases (anchoring, confirmation, change blindness) amplify. Eight hybrid design patterns address these failures: Generative UI, Infinite Canvas, Deictic Interaction, State Rail, Ghost Layers, Mise en Place, Semantic Zoom, and Probabilistic UI. Each pattern targets specific cognitive bottlenecks while preserving natural language for intent specification and synthesis. Well-scaffolded conversational systems that encode expert priors may reduce load for guided tasks; the framework applies most strongly to open-ended exploration. The paper concludes with falsifiable hypotheses and experimental paradigms for empirical validation.

LLMs are multilingual by training, yet their lingua franca is often English, reflecting English language dominance in pretraining. Other languages remain in parametric memory but are systematically suppressed. We argue that language defaultness is governed by a sparse, low-rank control circuit, language neurons, that can be mechanistically isolated and safely steered. We introduce Neural FOXP2, that makes a chosen language (Hindi or Spanish) primary in a model by steering language-specific neurons. Neural FOXP2 proceeds in three stages: (i) Localize: We train per-layer SAEs so each activation decomposes into a small set of active feature components. For every feature, we quantify English vs. Hindi/Spanish selectivity overall logit-mass lift toward the target-language token set. Tracing the top-ranked features back to their strongest contributing units yields a compact language-neuron set. (ii) Steering directions: We localize controllable language-shift geometry via a spectral low-rank analysis. For each layer, we build English to target activation-difference matrices and perform layerwise SVD to extract the dominant singular directions governing language change. The eigengap and effective-rank spectra identify a compact steering subspace and an empirically chosen intervention window (where these directions are strongest and most stable). (iii) Steer: We apply a signed, sparse activation shift targeted to the language neurons. Concretely, within low to mid layers we add a positive steering along the target-language dominant directions and a compensating negative shift toward the null space for the English neurons, yielding controllable target-language defaultness.

Cold-start exploration is a core challenge in large-scale recommender systems: new or data-sparse items must receive traffic to estimate value, but over-exploration harms users and wastes impressions. In practice, Thompson Sampling (TS) is often initialized with a uniform Beta(1,1) prior, implicitly assuming a 50% success rate for unseen items. When true base rates are far lower, this optimistic prior systematically over-allocates to weak items. The impact is amplified by batched policy updates and pipeline latency: for hours, newly launched items can remain effectively "no data," so the prior dominates allocation before feedback is incorporated. We propose Dynamic Prior Thompson Sampling, a prior design that directly controls the probability that a new arm outcompetes the incumbent winner. Our key contribution is a closed-form quadratic solution for the prior mean that enforces P(X_j > Y_k) = epsilon at introduction time, making exploration intensity predictable and tunable while preserving TS Bayesian updates. Across Monte Carlo validation, offline batched simulations, and a large-scale online experiment on a thumbnail personalization system serving millions of users, dynamic priors deliver precise exploration control and improved efficiency versus a uniform-prior baseline.

Modern large language models are increasingly deployed under compute and memory constraints, making flexible control of model capacity a central challenge. While sparse and low-rank structures naturally trade off capacity and performance, existing approaches often rely on heuristic designs that ignore layer and matrix heterogeneity or require model-specific architectural modifications. We propose SALAAD, a plug-and-play framework applicable to different model architectures that induces sparse and low-rank structures during training. By formulating structured weight learning under an augmented Lagrangian framework and introducing an adaptive controller that dynamically balances the training loss and structural constraints, SALAAD preserves the stability of standard training dynamics while enabling explicit control over the evolution of effective model capacity during training. Experiments across model scales show that SALAAD substantially reduces memory consumption during deployment while achieving performance comparable to ad-hoc methods. Moreover, a single training run yields a continuous spectrum of model capacities, enabling smooth and elastic deployment across diverse memory budgets without the need for retraining.

Contaminated mixture of experts (MoE) is motivated by transfer learning methods where a pre-trained model, acting as a frozen expert, is integrated with an adapter model, functioning as a trainable expert, in order to learn a new task. Despite recent efforts to analyze the convergence behavior of parameter estimation in this model, there are still two unresolved problems in the literature. First, the contaminated MoE model has been studied solely in regression settings, while its theoretical foundation in classification settings remains absent. Second, previous works on MoE models for classification capture pointwise convergence rates for parameter estimation without any guaranty of minimax optimality. In this work, we close these gaps by performing, for the first time, the convergence analysis of a contaminated mixture of multinomial logistic experts with homogeneous and heterogeneous structures, respectively. In each regime, we characterize uniform convergence rates for estimating parameters under challenging settings where ground-truth parameters vary with the sample size. Furthermore, we also establish corresponding minimax lower bounds to ensure that these rates are minimax optimal. Notably, our theories offer an important insight into the design of contaminated MoE, that is, expert heterogeneity yields faster parameter estimation rates and, therefore, is more sample-efficient than expert homogeneity.

Leveraging pre-trained 2D image representations in behavior cloning policies has achieved great success and has become a standard approach for robotic manipulation. However, such representations fail to capture the 3D spatial information about objects and scenes that is essential for precise manipulation. In this work, we introduce Contrastive Learning for 3D Multi-View Action-Conditioned Robotic Manipulation Pretraining (CLAMP), a novel 3D pre-training framework that utilizes point clouds and robot actions. From the merged point cloud computed from RGB-D images and camera extrinsics, we re-render multi-view four-channel image observations with depth and 3D coordinates, including dynamic wrist views, to provide clearer views of target objects for high-precision manipulation tasks. The pre-trained encoders learn to associate the 3D geometric and positional information of objects with robot action patterns via contrastive learning on large-scale simulated robot trajectories. During encoder pre-training, we pre-train a Diffusion Policy to initialize the policy weights for fine-tuning, which is essential for improving fine-tuning sample efficiency and performance. After pre-training, we fine-tune the policy on a limited amount of task demonstrations using the learned image and action representations. We demonstrate that this pre-training and fine-tuning design substantially improves learning efficiency and policy performance on unseen tasks. Furthermore, we show that CLAMP outperforms state-of-the-art baselines across six simulated tasks and five real-world tasks.

The Model Context Protocol (MCP) is rapidly becoming the standard interface for Large Language Models (LLMs) to discover and invoke external tools. However, existing evaluations often fail to capture the complexity of real-world scenarios, relying on restricted toolsets, simplistic workflows, or subjective LLM-as-a-judge metrics. We introduce MCP-Atlas, a large-scale benchmark for evaluating tool-use competency, comprising 36 real MCP servers and 220 tools. It includes 1,000 tasks designed to assess tool-use competency in realistic, multi-step workflows. Tasks use natural language prompts that avoid naming specific tools or servers, requiring agents to identify and orchestrate 3-6 tool calls across multiple servers. We score tasks using a claims-based rubric that awards partial credit based on the factual claims satisfied in the model's final answer, complemented by internal diagnostics on tool discovery, parameterization, syntax, error recovery, and efficiency. Evaluation results on frontier models reveal that top models achieve pass rates exceeding 50%, with primary failures arising from inadequate tool usage and task understanding. We release the task schema, containerized harness, and a 500-task public subset of the benchmark dataset to facilitate reproducible comparisons and advance the development of robust, tool-augmented agents.

Large language model reasoning is often treated as a monolithic capability, relying on binary preference supervision that fails to capture partial progress or fine-grained reasoning quality. We introduce Continuous Utility Direct Preference Optimization (CU-DPO), a framework that aligns models to a portfolio of prompt-based cognitive strategies by replacing binary labels with continuous scores that capture fine-grained reasoning quality. We prove that learning with K strategies yields a Theta(K log K) improvement in sample complexity over binary preferences, and that DPO converges to the entropy-regularized utility-maximizing policy. To exploit this signal, we propose a two-stage training pipeline: (i) strategy selection, which optimizes the model to choose the best strategy for a given problem via best-vs-all comparisons, and (ii) execution refinement, which trains the model to correctly execute the selected strategy using margin-stratified pairs. On mathematical reasoning benchmarks, CU-DPO improves strategy selection accuracy from 35-46 percent to 68-78 percent across seven base models, yielding consistent downstream reasoning gains of up to 6.6 points on in-distribution datasets with effective transfer to out-of-distribution tasks.

Humans learn abstractions and use them to plan efficiently to quickly generalize across tasks -- an ability that remains challenging for state-of-the-art large language model (LLM) agents and deep reinforcement learning (RL) systems. Inspired by the cognitive science of how people form abstractions and intuitive theories of their world knowledge, Theory-Based RL (TBRL) systems, such as TheoryCoder, exhibit strong generalization through effective use of abstractions. However, they heavily rely on human-provided abstractions and sidestep the abstraction-learning problem. We introduce TheoryCoder-2, a new TBRL agent that leverages LLMs' in-context learning ability to actively learn reusable abstractions rather than relying on hand-specified ones, by synthesizing abstractions from experience and integrating them into a hierarchical planning process. We conduct experiments on diverse environments, including BabyAI, Minihack and VGDL games like Sokoban. We find that TheoryCoder-2 is significantly more sample-efficient than baseline LLM agents augmented with classical planning domain construction, reasoning-based planning, and prior program-synthesis agents such as WorldCoder. TheoryCoder-2 is able to solve complex tasks that the baselines fail, while only requiring minimal human prompts, unlike prior TBRL systems.

Training LLMs to think and reason for longer has become a key ingredient in building state-of-the-art models that can solve complex problems previously out of reach. Recent efforts pursue this in different ways, such as RL fine-tuning to elicit long CoT or scaling latent reasoning through architectural recurrence. This makes reasoning length an important scaling knob. In this work, we identify a novel phenomenon (both theoretically and experimentally): under outcome-only supervision, out-of-distribution (OOD) performance can continue improving as training-time reasoning length (e.g., the token budget in RL, or the loop count in looped Transformers) increases, even after in-distribution (ID) performance has saturated. This suggests that robustness may require a larger budget than ID validation alone would indicate. We provide theoretical explanations via two mechanisms: (i) self-iteration can induce a stronger inductive bias in the hypothesis class, reshaping ID-optimal solutions in ways that improve OOD generalization; and (ii) when shortcut solutions that work for ID samples but not for OOD samples persist in the hypothesis class, regularization can reduce the learned solution's reliance on these shortcuts as the number of self-iterations increases. We complement the theory with empirical evidence from two realizations of scaling training-time reasoning length: increasing the number of loops in looped Transformers on a synthetic task, and increasing token budgets during RL fine-tuning of LLMs on mathematical reasoning.

Sparse auto-encoders (SAEs) have re-emerged as a prominent method for mechanistic interpretability, yet they face two significant challenges: the non-smoothness of the $L_1$ penalty, which hinders reconstruction and scalability, and a lack of alignment between learned features and human semantics. In this paper, we address these limitations by adapting unconstrained feature models-a mathematical framework from neural collapse theory-and by supervising the task. We supervise (decoder-only) SAEs to reconstruct feature vectors by jointly learning sparse concept embeddings and decoder weights. Validated on Stable Diffusion 3.5, our approach demonstrates compositional generalization, successfully reconstructing images with concept combinations unseen during training, and enabling feature-level intervention for semantic image editing without prompt modification.

Optimal feedback control with implicit Hamiltonians poses a fundamental challenge for learning-based value function methods due to the absence of closed-form optimal control laws. Recent work~\cite{gelphman2025end} introduced an implicit deep learning approach using Jacobian-Free Backpropagation (JFB) to address this setting, but only established sample-wise descent guarantees. In this paper, we establish convergence guarantees for JFB in the stochastic minibatch setting, showing that the resulting updates converge to stationary points of the expected optimal control objective. We further demonstrate scalability on substantially higher-dimensional problems, including multi-agent optimal consumption and swarm-based quadrotor and bicycle control. Together, our results provide both theoretical justification and empirical evidence for using JFB in high-dimensional optimal control with implicit Hamiltonians.

Early Classification of Time Series (ECTS) addresses decision-making problems in which predictions must be made as early as possible while maintaining high accuracy. Most existing ECTS methods assume that the time-dependent decision costs governing the learning objective are known, fixed, and correctly specified. In practice, however, these costs are often uncertain and may change over time, leading to mismatches between training-time and deployment-time objectives. In this paper, we study ECTS under two practically relevant forms of cost non-stationarity: drift in the balance between misclassification and decision delay costs, and stochastic realizations of decision costs that deviate from the nominal training-time model. To address these challenges, we revisit representative ECTS approaches and adapt them to an online learning setting. Focusing on separable methods, we update only the triggering model during deployment, while keeping the classifier fixed. We propose several online adaptations and baselines, including bandit-based and RL-based approaches, and conduct controlled experiments on synthetic data to systematically evaluate robustness under cost non-stationarity. Our results demonstrate that online learning can effectively improve the robustness of ECTS methods to cost drift, with RL-based strategies exhibiting strong and stable performance across varying cost regimes.

We present a lightweight multimodal baseline for emotion recognition in conversations using the SemEval-2024 Task 3 dataset built from the sitcom Friends. The goal of this report is not to propose a novel state-of-the-art method, but to document an accessible reference implementation that combines (i) a transformer-based text classifier and (ii) a self-supervised speech representation model, with a simple late-fusion ensemble. We report the baseline setup and empirical results obtained under a limited training protocol, highlighting when multimodal fusion improves over unimodal models. This preprint is provided for transparency and to support future, more rigorous comparisons.

Sentence-level human value detection is typically framed as multi-label classification over Schwartz values, but it remains unclear whether Schwartz higher-order (HO) categories provide usable structure. We study this under a strict compute-frugal budget (single 8 GB GPU) on ValueEval'24 / ValuesML (74K English sentences). We compare (i) direct supervised transformers, (ii) HO$\rightarrow$values pipelines that enforce the hierarchy with hard masks, and (iii) Presence$\rightarrow$HO$\rightarrow$values cascades, alongside low-cost add-ons (lexica, short context, topics), label-wise threshold tuning, small instruction-tuned LLM baselines ($\le$10B), QLoRA, and simple ensembles. HO categories are learnable from single sentences (e.g., the easiest bipolar pair reaches Macro-$F_1\approx0.58$), but hard hierarchical gating is not a reliable win: it often reduces end-task Macro-$F_1$ via error compounding and recall suppression. In contrast, label-wise threshold tuning is a high-leverage knob (up to $+0.05$ Macro-$F_1$), and small transformer ensembles provide the most consistent additional gains (up to $+0.02$ Macro-$F_1$). Small LLMs lag behind supervised encoders as stand-alone systems, yet can contribute complementary errors in cross-family ensembles. Overall, HO structure is useful descriptively, but enforcing it with hard gates hurts sentence-level value detection; robust improvements come from calibration and lightweight ensembling.

Collaborative learning among LLM-based agents under federated learning faces challenges, including communication costs, heterogeneity in data, and tool-usage, limiting their effectiveness. We introduce Synapse, a framework that trains a shared global knowledge model of tool-usage behavior. Client agents with fixed LLMs learn tool-usage patterns locally, and transmit artifacts for federated aggregation through coordinators. A global tool compendium is updated and redistributed, enabling convergence toward stable tool selection. Synapse uses templated representations, embedding retrieval with LLM reranking, and adaptive masking to maintain utility while limiting information leakage. The framework supports heterogeneous data and quantifies performance improvements. Results show that Synapse improves tool-usage effectiveness and reduces communication overhead compared with weight or prompt-sharing approaches in multi-agent LLM systems.

Deep learning has revolutionized medical image analysis, playing a vital role in modern clinical applications. However, the deployment of large-scale models in real-world clinical settings remains challenging due to high computational costs, latency constraints, and patient data privacy concerns associated with cloud-based processing. To address these bottlenecks, this review provides a comprehensive synthesis of efficient and lightweight deep learning architectures specifically tailored for the medical domain. We categorize the landscape of modern efficient models into three primary streams: Convolutional Neural Networks (CNNs), Lightweight Transformers, and emerging Linear Complexity Models. Furthermore, we examine key model compression strategies (including pruning, quantization, knowledge distillation, and low-rank factorization) and evaluate their efficacy in maintaining diagnostic performance while reducing hardware requirements. By identifying current limitations and discussing the transition toward on-device intelligence, this review serves as a roadmap for researchers and practitioners aiming to bridge the gap between high-performance AI and resource-constrained clinical environments.

PyGALAX is a Python package for geospatial analysis that integrates automated machine learning (AutoML) and explainable artificial intelligence (XAI) techniques to analyze spatial heterogeneity in both regression and classification tasks. It automatically selects and optimizes machine learning models for different geographic locations and contexts while maintaining interpretability through SHAP (SHapley Additive exPlanations) analysis. PyGALAX builds upon and improves the GALAX framework (Geospatial Analysis Leveraging AutoML and eXplainable AI), which has proven to outperform traditional geographically weighted regression (GWR) methods. Critical enhancements in PyGALAX from the original GALAX framework include automatic bandwidth selection and flexible kernel function selection, providing greater flexibility and robustness for spatial modeling across diverse datasets and research questions. PyGALAX not only inherits all the functionalities of the original GALAX framework but also packages them into an accessible, reproducible, and easily deployable Python toolkit while providing additional options for spatial modeling. It effectively addresses spatial non-stationarity and generates transparent insights into complex spatial relationships at both global and local scales, making advanced geospatial machine learning methods accessible to researchers and practitioners in geography, urban planning, environmental science, and related fields.

Large language models often hallucinate with high confidence on "random facts" that lack inferable patterns. We formalize the memorization of such facts as a membership testing problem, unifying the discrete error metrics of Bloom filters with the continuous log-loss of LLMs. By analyzing this problem in the regime where facts are sparse in the universe of plausible claims, we establish a rate-distortion theorem: the optimal memory efficiency is characterized by the minimum KL divergence between score distributions on facts and non-facts. This theoretical framework provides a distinctive explanation for hallucination: even with optimal training, perfect data, and a simplified "closed world" setting, the information-theoretically optimal strategy under limited capacity is not to abstain or forget, but to assign high confidence to some non-facts, resulting in hallucination. We validate this theory empirically on synthetic data, showing that hallucinations persist as a natural consequence of lossy compression.

E-commerce recommendation and search commonly rely on sparse keyword matching (e.g., BM25), which breaks down under vocabulary mismatch when user intent has limited lexical overlap with product metadata. We cast content-based recommendation as recommendation-as-retrieval: given a natural-language intent signal (a query or review), retrieve the top-K most relevant items from a large catalog via semantic similarity. We present a scalable dense retrieval system based on a two-tower bi-encoder, fine-tuned on the Amazon Reviews 2023 (Fashion) subset using supervised contrastive learning with Multiple Negatives Ranking Loss. We construct training pairs from review text (as a query proxy) and item metadata (as the positive document) and fine-tune on 50,000 sampled interactions with a maximum sequence length of 500 tokens. For efficient serving, we combine FAISS HNSW indexing with an ONNX Runtime inference pipeline using INT8 dynamic quantization. On a review-to-title benchmark over 826,402 catalog items, our approach improves Recall@10 from 0.26 (BM25) to 0.66, while meeting practical latency and model-size constraints: 6.1 ms median CPU inference latency (batch size 1) and a 4x reduction in model size. Overall, we provide an end-to-end, reproducible blueprint for taking domain-adapted dense retrieval from offline training to CPU-efficient serving at catalog scale.

We present a fast sparse matrix permutation algorithm tailored to linear systems arising from triangle meshes. Our approach produces nested-dissection-style permutations while significantly reducing permutation runtime overhead. Rather than enforcing strict balance and separator optimality, the algorithm deliberately relaxes these design decisions to favor fast partitioning and efficient elimination-tree construction. Our method decomposes permutation into patch-level local orderings and a compact quotient-graph ordering of separators, preserving the essential structure required by sparse Cholesky factorization while avoiding its most expensive components. We integrate our algorithm into vendor-maintained sparse Cholesky solvers on both CPUs and GPUs. Across a range of graphics applications, including single factorizations, repeated factorizations, our method reduces permutation time and improves the sparse Cholesky solve performance by up to 6.27x.

Photonic in-memory computing is a high-speed, low-energy alternative to traditional transistor-based digital computing that utilizes high photonic operating frequencies and bandwidths. In this work, we develop a comprehensive system-level performance model for photonic in-memory computing, capturing the effects of key latency sources such as external memory access and opto-electronic conversion. We perform algorithm-to-hardware mapping across a range of workloads, including the Sod shock tube problem, Matricized Tensor Times Khatri-Rao Product (MTTKRP), and the Vlasov-Maxwell equation, to evaluate how the latencies impact real-world high-performance computing workloads. Our performance model shows that, while accounting for system overheads, a compact 1x256 bit single-wavelength photonic SRAM array, fabricated using the standard silicon photonics process by GlobalFoundries, sustains up to 1.5 TOPS, 0.9 TOPS, and 1.3 TOPS on the Sod shock tube problem, MTTKRP, and the Vlasov-Maxwell equation with an average energy efficiency of 2.5 TOPS/W.

The advent of the web has led to a paradigm shift in the financial relations, with the real-time dissemination of news, social discourse, and financial filings contributing significantly to the reshaping of financial forecasting. The existing methods rely on establishing relations a priori, i.e. predefining graphs to capture inter-stock relationships. However, the stock-related web signals are characterised by high levels of noise, asynchrony, and challenging to obtain, resulting in poor generalisability and non-alignment between the predefined graphs and the downstream tasks. To address this, we propose GAPNet, a Graph Adaptation Plug-in Network that jointly learns task-specific topology and representations in an end-to-end manner. GAPNet attaches to existing pairwise graph or hypergraph backbone models, enabling the dynamic adaptation and rewiring of edge topologies via two complementary components: a Spatial Perception Layer that captures short-term co-movements across assets, and a Temporal Perception Layer that maintains long-term dependency under distribution shift. Across two real-world stock datasets, GAPNet has been shown to consistently enhance the profitability and stability in comparision to the state-of-the-art models, yielding annualised cumulative returns of up to 0.47 for RT-GCN and 0.63 for CI-STHPAN, with peak Sharpe Ratio of 2.20 and 2.12 respectively. The plug-and-play design of GAPNet ensures its broad applicability to diverse GNN-based architectures. Our results underscore that jointly learning graph structures and representations is essential for task-specific relational modeling.

Most existing language model agentic systems today are built and optimized for large language models (e.g., GPT, Claude, Gemini) via API calls. While powerful, this approach faces several limitations including high token costs and privacy concerns for sensitive applications. We introduce effGen, an open-source agentic framework optimized for small language models (SLMs) that enables effective, efficient, and secure local deployment (pip install effgen). effGen makes four major contributions: (1) Enhanced tool-calling with prompt optimization that compresses contexts by 70-80% while preserving task semantics, (2) Intelligent task decomposition that breaks complex queries into parallel or sequential subtasks based on dependencies, (3) Complexity-based routing using five factors to make smart pre-execution decisions, and (4) Unified memory system combining short-term, long-term, and vector-based storage. Additionally, effGen unifies multiple agent protocols (MCP, A2A, ACP) for cross-protocol communication. Results on 13 benchmarks show effGen outperforms LangChain, AutoGen, and Smolagents with higher success rates, faster execution, and lower memory. Our results reveal that prompt optimization and complexity routing have complementary scaling behavior: optimization benefits SLMs more (11.2% gain at 1.5B vs 2.4% at 32B), while routing benefits large models more (3.6% at 1.5B vs 7.9% at 32B), providing consistent gains across all scales when combined. effGen (https://effgen.org/) is released under the MIT License, ensuring broad accessibility for research and commercial use. Our framework code is publicly available at https://github.com/ctrl-gaurav/effGen.

Diffusion policies are a powerful paradigm for robotic control, but fine-tuning them with human preferences is fundamentally challenged by the multi-step structure of the denoising process. To overcome this, we introduce a Unified Markov Decision Process (MDP) formulation that coherently integrates the diffusion denoising chain with environmental dynamics, enabling reward-free Direct Preference Optimization (DPO) for diffusion policies. Building on this formulation, we propose RoDiF (Robust Direct Fine-Tuning), a method that explicitly addresses corrupted human preferences. RoDiF reinterprets the DPO objective through a geometric hypothesis-cutting perspective and employs a conservative cutting strategy to achieve robustness without assuming any specific noise distribution. Extensive experiments on long-horizon manipulation tasks show that RoDiF consistently outperforms state-of-the-art baselines, effectively steering pretrained diffusion policies of diverse architectures to human-preferred modes, while maintaining strong performance even under 30% corrupted preference labels.

Informative data selection is a key requirement for large language models (LLMs) to minimize the amount of data required for fine-tuning, network distillation, and token pruning, enabling fast and efficient deployment, especially under computational and communication constraints. Traditional subset selection methods, including those based on Determinantal Point Processes (DPP), focus on maximizing diversity but assume that selected data batches are always available error-free. This presumption prohibits their use under partial storage outage, imperfect communication, and stochastic access failures. Furthermore, we show that the original formulation collapses under such conditions. To address this gap, we introduce ProbDPP, a novel reliability-aware implementation of k-DPP that accounts for probabilistic data access by recasting the objective function with a regularization term that remains well-posed and decomposes into a geometric diversity term and unreliability cost. The resulting objective facilitates robust selection of diverse data batches under uncertainty. Furthermore, we frame this reliability-aware diversity maximization as a combinatorial semi-bandit problem and propose a UCB-style algorithm to efficiently learn the unknown reliability online. Theoretical analysis provides regret bounds for the proposed approach, ensuring performance guarantees.

Neural operators have shown promise in learning solution maps of partial differential equations (PDEs), but they often struggle to generalize when test inputs lie outside the training distribution, such as novel initial conditions, unseen PDE coefficients or unseen physics. Prior works address this limitation with large-scale multiple physics pretraining followed by fine-tuning, but this still requires examples from the new dynamics, falling short of true zero-shot generalization. In this work, we propose a method to enhance generalization at test time, i.e., without modifying pretrained weights. Building on DISCO, which provides a dictionary of neural operators trained across different dynamics, we introduce a neural operator splitting strategy that, at test time, searches over compositions of training operators to approximate unseen dynamics. On challenging out-of-distribution tasks including parameter extrapolation and novel combinations of physics phenomena, our approach achieves state-of-the-art zero-shot generalization results, while being able to recover the underlying PDE parameters. These results underscore test-time computation as a key avenue for building flexible, compositional, and generalizable neural operators.

Despite impressive results from recent text-to-image models like FLUX, visual and anatomical artifacts remain a significant hurdle for practical and professional use. Existing methods for artifact reduction, typically work in a post-hoc manner, consequently failing to intervene effectively during the core image formation process. Notably, current techniques require problematic and invasive modifications to the model weights, or depend on a computationally expensive and time-consuming process of regional refinement. To address these limitations, we propose DIAMOND, a training-free method that applies trajectory correction to mitigate artifacts during inference. By reconstructing an estimate of the clean sample at every step of the generative trajectory, DIAMOND actively steers the generation process away from latent states that lead to artifacts. Furthermore, we extend the proposed method to standard Diffusion Models, demonstrating that DIAMOND provides a robust, zero-shot path to high-fidelity, artifact-free image synthesis without the need for additional training or weight modifications in modern generative architectures. Code is available at https://gmum.github.io/DIAMOND/

Legal Statute Identification (LSI) for a given situation is one of the most fundamental tasks in Legal NLP. This task has traditionally been modeled using facts from court judgments as input queries, due to their abundance. However, in practical settings, the input queries are likely to be informal and asked by laypersons, or non-professionals. While a few laypeople LSI datasets exist, there has been little research to explore the differences between court and laypeople data for LSI. In this work, we create ILSIC, a corpus of laypeople queries covering 500+ statutes from Indian law. Additionally, the corpus also contains court case judgements to enable researchers to effectively compare between court and laypeople data for LSI. We conducted extensive experiments on our corpus, including benchmarking over the laypeople dataset using zero and few-shot inference, retrieval-augmented generation and supervised fine-tuning. We observe that models trained purely on court judgements are ineffective during test on laypeople queries, while transfer learning from court to laypeople data can be beneficial in certain scenarios. We also conducted fine-grained analyses of our results in terms of categories of queries and frequency of statutes.

Among parallel decoding paradigms, diffusion large language models (dLLMs) have emerged as a promising candidate that balances generation quality and throughput. However, their integration with Mixture-of-Experts (MoE) architectures is constrained by an expert explosion: as the number of tokens generated in parallel increases, the number of distinct experts activated grows nearly linearly. This results in substantial memory traffic that pushes inference into a memory-bound regime, negating the efficiency gains of both MoE and parallel decoding. To address this challenge, we propose Dynamic Expert Sharing (DES), a novel technique that shifts MoE optimization from token-centric pruning and conventional expert skipping methods to sequence-level coreset selection. To maximize expert reuse, DES identifies a compact, high-utility set of experts to satisfy the requirements of an entire parallel decoding block. We introduce two innovative selection strategies: (1) Intra-Sequence Sharing (DES-Seq), which adapts optimal allocation to the sequence level, and (2) Saliency-Aware Voting (DES-Vote), a novel mechanism that allows tokens to collectively elect a coreset based on aggregated router weights. Extensive experiments on MoE dLLMs demonstrate that DES reduces unique expert activations by over 55% and latency by up to 38%, while retaining 99% of vanilla accuracy, effectively decoupling memory overhead from the degree of parallelism.

Given the query, key and value matrices $Q, K, V\in \mathbb{R}^{n\times d}$, the attention module is defined as $\mathrm{Att}(Q, K, V)=D^{-1}AV$ where $A=\exp(QK^\top/\sqrt{d})$ with $\exp(\cdot)$ applied entrywise, $D=\mathrm{diag}(A{\bf 1}_n)$. The attention module is the backbone of modern transformers and large language models, but explicitly forming the softmax matrix $D^{-1}A$ incurs $Ω(n^2)$ time, motivating numerous approximation schemes that reduce runtime to $\widetilde O(nd)$ via sparsity or low-rank factorization. We propose a quantum data structure that approximates any row of $\mathrm{Att}(Q, K, V)$ using only row queries to $Q, K, V$. Our algorithm preprocesses these matrices in $\widetilde{O}\left( ε^{-1} n^{0.5} \left( s_λ^{2.5} + s_λ^{1.5} d + α^{0.5} d \right) \right)$ time, where $ε$ is the target accuracy, $s_λ$ is the $λ$-statistical dimension of the exponential kernel defined by $Q$ and $K$, and $α$ measures the row distortion of $V$ that is at most $d/{\rm srank}(V)$, the stable rank of $V$. Each row query can be answered in $\widetilde{O}(s_λ^2 + s_λd)$ time. To our knowledge, this is the first quantum data structure that approximates rows of the attention matrix in sublinear time with respect to $n$. Our approach relies on a quantum Nyström approximation of the exponential kernel, quantum multivariate mean estimation for computing $D$, and quantum leverage score sampling for the multiplication with $V$.

We study solution learning for heat-based equations in self-similar variables (SSV). We develop an SSV training framework compatible with standard neural-operator training. We instantiate this framework on the two-dimensional incompressible Navier-Stokes equations and the one-dimensional viscous Burgers equation, and perform controlled comparisons between models trained in physical coordinates and in the corresponding self-similar coordinates using two simple fully connected architectures (standard multilayer perceptrons and a factorized fully connected network). Across both systems and both architectures, SSV-trained networks consistently deliver substantially more accurate and stable extrapolation beyond the training window and better capture qualitative long-time trends. These results suggest that self-similar coordinates provide a mathematically motivated inductive bias for learning the long-time dynamics of heat-based equations.

Large language models (LLMs) have shown promising self-correction abilities, where iterative refinement improves the quality of generated responses. However, most existing approaches operate at the level of output critique, patching surface errors while often failing to correct deeper reasoning flaws. We propose SELF-THOUGHT, a framework that introduces an intermediate step of task abstraction before solution refinement. Given an input and an initial response, the model first distills the task into a structured template that captures key variables, constraints, and problem structure. This abstraction then guides solution instantiation, grounding subsequent responses in a clearer understanding of the task and reducing error propagation. Crucially, we show that these abstractions can be transferred across models: templates generated by larger models can serve as structured guides for smaller LLMs, which typically struggle with intrinsic self-correction. By reusing distilled task structures, smaller models achieve more reliable refinements without heavy fine-tuning or reliance on external verifiers. Experiments across diverse reasoning tasks demonstrate that SELF-THOUGHT improves accuracy, robustness, and generalization for both large and small models, offering a scalable path toward more reliable self-correcting language systems.

Conditional flow matching (CFM) stands out as an efficient, simulation-free approach for training flow-based generative models, achieving remarkable performance for data generation. However, CFM is insufficient to ensure accuracy in learning probability paths. In this paper, we introduce a new partial differential equation characterization for the error between the learned and exact probability paths, along with its solution. We show that the total variation gap between the two probability paths is bounded above by a combination of the CFM loss and an associated divergence loss. This theoretical insight leads to the design of a new objective function that simultaneously matches the flow and its divergence. Our new approach improves the performance of the flow-based generative model by a noticeable margin without sacrificing generation efficiency. We showcase the advantages of this enhanced training approach over CFM on several important benchmark tasks, including generative modeling for dynamical systems, DNA sequences, and videos. Code is available at \href{https://github.com/Utah-Math-Data-Science/Flow_Div_Matching}{Utah-Math-Data-Science}.

The Controller Area Network (CAN) bus provides a rich source of vehicular signals increasingly leveraged for applications in automotive and auto insurance domains, including collision detection, predictive maintenance, and driver risk modeling. Despite this potential, existing pipelines largely train isolated task-specific models on raw CAN data, with only limited efforts exploring decoded signals. Such fragmentation prevents shared representation learning and limits cross-task generalization. By contrast, natural language processing (NLP) and computer vision (CV) have been transformed by the foundation model paradigm: large-scale pretraining followed by task-specific adaptation. In this work, we introduce the foundation CAN model that demonstrates multi-objective downstream generalization using a single pretrained backbone. Our approach treats CAN data as a language: we pretrain on large-scale, unlabeled decoded CAN signals and fine-tune across heterogeneous auto insurance tasks. To enable this, we propose a unified tokenization scheme for mixed discrete-continuous signals and address challenges of temporal complexity and trip-specific variability. Our results show that one pretrained CAN model can adapt effectively to diverse predictive tasks, validating that the foundation modeling paradigm, proven in NLP and CV, also holds for CAN data. This establishes a new direction for generalizable representation learning in automotive AI.

Graph neural networks (GNNs) have emerged as powerful tools for learning protein structures by capturing spatial relationships at the residue level. However, existing GNN-based methods often face challenges in learning multiscale representations and modeling long-range dependencies efficiently. In this work, we propose an efficient multiscale graph-based learning framework tailored to proteins. Our proposed framework contains two crucial components: (1) It constructs a hierarchical graph representation comprising a collection of fine-grained subgraphs, each corresponding to a secondary structure motif (e.g., $α$-helices, $β$-strands, loops), and a single coarse-grained graph that connects these motifs based on their spatial arrangement and relative orientation. (2) It employs two GNNs for feature learning: the first operates within individual secondary motifs to capture local interactions, and the second models higher-level structural relationships across motifs. Our modular framework allows a flexible choice of GNN in each stage. Theoretically, we show that our hierarchical framework preserves the desired maximal expressiveness, ensuring no loss of critical structural information. Empirically, we demonstrate that integrating baseline GNNs into our multiscale framework remarkably improves prediction accuracy and reduces computational cost across various benchmarks.

Modern transformer attention is internally multi-agent -- heads compete and coordinate -- yet we train it as if it were a monolithic optimizer. We formalize this gap: cross-entropy training induces an implicit potential game among heads, and gradient descent converges to Nash equilibria with potentially unbounded inefficiency due to unpriced externalities (redundancy, correlated errors). Our main result bounds the Price of Anarchy by $Γ(G)$, the off-diagonal mass of a head interaction matrix capturing weight and gradient coupling. Under mild smoothness assumptions, we prove that both \emph{excess hallucination probability} and \emph{excess head redundancy} scale with PoA, unifying two distinct failure modes into a single mechanism. The bound is prescriptive: regularization that reduces $Γ(G)$ provably tightens PoA. We instantiate this as GAME-LoRA, combining Barlow Twins decorrelation with log-determinant coordination pressure. Experiments validate the theory: $Γ(G)$ predicts hallucination ($p{<}0.05$), emergent coalitions exhibit selective coordination, and GAME-LoRA achieves up to 18\% hallucination reduction (8\% average) with no knowledge degradation -- a Pareto improvement inaccessible to methods ignoring the game structure.

Research on adversarial robustness in language models is currently fragmented across applications and attacks, obscuring shared vulnerabilities. In this work, we propose unifying the study of adversarial robustness in text scoring models spanning dense retrievers, rerankers, and reward models. This motivates adapting both attacks and adversarial training methods across model roles. Unlike open-ended generation, text scoring failures are directly testable: an attack succeeds when an irrelevant or rejected text outscores a relevant or chosen one. Using this principled lens of text scoring, we demonstrate that current adversarial training formulations for language models are often short-sighted, failing to effectively generalize across attacks. To address this, we introduce multiple adversarial training methods for text scoring models and show that combining complementary training methods can yield strong robustness while also improving task effectiveness. We also highlight the practical value of our approach for RLHF, showing that our adversarially trained reward models mitigate reward hacking and support the training of better-aligned LLMs. We provide our code and models for further study.

AI systems increasingly produce fluent, correct, end-to-end outcomes. Over time, this erodes users' ability to explain, verify, or intervene. We define this divergence as the Capability-Comprehension Gap: a decoupling where assisted performance improves while users' internal models deteriorate. This paper argues that prevailing approaches to transparency, user control, literacy, and governance do not define the foundational understanding humans must retain for oversight under sustained AI delegation. To formalize this, we define the Cognitive Integrity Threshold (CIT) as the minimum comprehension required to preserve oversight, autonomy, and accountable participation under AI assistance. CIT does not require full reasoning reconstruction, nor does it constrain automation. It identifies the threshold beyond which oversight becomes procedural and contestability fails. We operatinalize CIT through three functional dimensions: (i) verification capacity, (ii) comprehension-preserving interaction, and (iii) institutional scaffolds for governance. This motivates a design and governance agenda that aligns human-AI interaction with cognitive sustainability in responsibility-critical settings.

Subtask distillation is an emerging paradigm in which compact, specialized models are extracted from large, general-purpose 'foundation models' for deployment in environments with limited resources or in standalone computer systems. Although distillation uses a teacher model, it still relies on a dataset that is often limited in size and may lack representativeness or exhibit spurious correlations. In this paper, we evaluate established distillation methods, as well as the recent SubDistill method, when using data with spurious correlations for distillation. As the strength of the correlations increases, we observe a widening gap between advanced methods, such as SubDistill, which remain fairly robust, and some baseline methods, which degrade to near-random performance. Overall, our study underscores the challenges of knowledge distillation when applied to imperfect, real-world datasets, particularly those with spurious correlations.

Modern AI agents increasingly combine conversational interaction with autonomous task execution, such as coding and web research, raising a natural question: what happens when an agent engaged in long-horizon tasks is subjected to user persuasion? We study how belief-level intervention can influence downstream task behavior, a phenomenon we name \emph{persuasion propagation}. We introduce a behavior-centered evaluation framework that distinguishes between persuasion applied during or prior to task execution. Across web research and coding tasks, we find that on-the-fly persuasion induces weak and inconsistent behavioral effects. In contrast, when the belief state is explicitly specified at task time, belief-prefilled agents conduct on average 26.9\% fewer searches and visit 16.9\% fewer unique sources than neutral-prefilled agents. These results suggest that persuasion, even in prior interaction, can affect the agent's behavior, motivating behavior-level evaluation in agentic systems.

Mean flow (MeanFlow) enables efficient, high-fidelity image generation, yet its single-function evaluation (1-NFE) generation often cannot yield compelling results. We address this issue by introducing RMFlow, an efficient multimodal generative model that integrates a coarse 1-NFE MeanFlow transport with a subsequent tailored noise-injection refinement step. RMFlow approximates the average velocity of the flow path using a neural network trained with a new loss function that balances minimizing the Wasserstein distance between probability paths and maximizing sample likelihood. RMFlow achieves near state-of-the-art results on text-to-image, context-to-molecule, and time-series generation using only 1-NFE, at a computational cost comparable to the baseline MeanFlows.

Large language models (LLMs) encode knowledge with varying degrees of confidence. When responding to queries, models face an inherent trade-off: they can generate responses that are less informative but highly factual, or more informative but potentially less accurate. Different applications demand different balances between informativeness and factuality. We introduce Factuality-Controlled Generation (FCG), a framework that enables users to specify factuality constraints alongside their queries. We propose to evaluate FCG performance on two dimensions: adherence to factuality constraints and response informativeness. We propose to train models on the FCG task using synthetic data, and show that our synthetic training significantly improves models' ability to both respect factuality requirements and maintain informativeness in their outputs.

Multimodal large language models (MLLMs) have shown remarkable capabilities, yet their performance is often capped by the coarse nature of existing alignment techniques. A critical bottleneck remains the lack of effective reward models (RMs): existing RMs are predominantly vision-centric, return opaque scalar scores, and rely on costly human annotations. We introduce \textbf{Omni-RRM}, the first open-source rubric-grounded reward model that produces structured, multi-dimension preference judgments with dimension-wise justifications across \textbf{text, image, video, and audio}. At the core of our approach is \textbf{Omni-Preference}, a large-scale dataset built via a fully automated pipeline: we synthesize candidate response pairs by contrasting models of different capabilities, and use strong teacher models to \emph{reconcile and filter} preferences while providing a modality-aware \emph{rubric-grounded rationale} for each pair. This eliminates the need for human-labeled training preferences. Omni-RRM is trained in two stages: supervised fine-tuning to learn the rubric-grounded outputs, followed by reinforcement learning (GRPO) to sharpen discrimination on difficult, low-contrast pairs. Comprehensive evaluations show that Omni-RRM achieves state-of-the-art accuracy on video (80.2\% on ShareGPT-V) and audio (66.8\% on Audio-HH-RLHF) benchmarks, and substantially outperforms existing open-source RMs on image tasks, with a 17.7\% absolute gain over its base model on overall accuracy. Omni-RRM also improves downstream performance via Best-of-$N$ selection and transfers to text-only preference benchmarks. Our data, code, and models are available at https://anonymous.4open.science/r/Omni-RRM-CC08.

Agentic reasoning enables large reasoning models (LRMs) to dynamically acquire external knowledge, but yet optimizing the retrieval process remains challenging due to the lack of dense, principled reward signals. In this paper, we introduce InfoReasoner, a unified framework that incentivizes effective information seeking via a synthetic semantic information gain reward. Theoretically, we redefine information gain as uncertainty reduction over the model's belief states, establishing guarantees, including non-negativity, telescoping additivity, and channel monotonicity. Practically, to enable scalable optimization without manual retrieval annotations, we propose an output-aware intrinsic estimator that computes information gain directly from the model's output distributions using semantic clustering via bidirectional textual entailment. This intrinsic reward guides the policy to maximize epistemic progress, enabling efficient training via Group Relative Policy Optimxization (GRPO). Experiments across seven question-answering benchmarks demonstrate that InfoReasoner consistently outperforms strong retrieval-augmented baselines, achieving up to 5.4% average accuracy improvement. Our work provides a theoretically grounded and scalable path toward agentic reasoning with retrieval.

Multivariate time series (MTS) imputation is often compromised by mismatch between observed and true data distributions -- a bias exacerbated by non-stationarity and systematic missingness. Standard methods that minimize reconstruction error or encourage distributional alignment risk overfitting these biased observations. We propose the Distributionally Robust Regularized Imputer Objective (DRIO), which jointly minimizes reconstruction error and the divergence between the imputer and a worst-case distribution within a Wasserstein ambiguity set. We derive a tractable dual formulation that reduces infinite-dimensional optimization over measures to adversarial search over sample trajectories, and propose an adversarial learning algorithm compatible with flexible deep learning backbones. Comprehensive experiments on diverse real-world datasets show DRIO consistently improves imputation under both missing-completely-at-random and missing-not-at-random settings, reaching Pareto-optimal trade-offs between reconstruction accuracy and distributional alignment.

Negabent Boolean functions are defined by having a flat magnitude spectrum under the nega-Hadamard transform. They exist in both even and odd dimensions, and the subclass of functions that are simultaneously bent and negabent (bent-negabent) has attracted interest due to the combined optimal periodic and negaperiodic spectral properties. In this work, we investigate how evolutionary algorithms can be used to evolve (bent-)negabent Boolean functions. Our experimental results indicate that evolutionary algorithms, especially genetic programming, are a suitable approach for evolving negabent Boolean functions, and we successfully evolve such functions in all dimensions we consider.

C/C++ is a prevalent programming language. Yet, it suffers from significant memory and thread-safety issues. Recent studies have explored automated translation of C/C++ to safer languages, such as Rust. However, these studies focused mostly on the correctness and safety of the translated code, which are indeed critical, but they left other important quality concerns (e.g., performance, robustness, and maintainability) largely unexplored. This work investigates strengths and weaknesses of three C-to-Rust translators, namely C2Rust (a transpiler), C2SaferRust (an LLM-guided transpiler), and TranslationGym (an LLM-based direct translation). We perform an in-depth quantitative and qualitative analysis of several important quality attributes for the translated Rust code of the popular GNU coreutils, using human-based translation as a baseline. To assess the internal and external quality of the Rust code, we: (i) apply Clippy, a rule-based state-of-the-practice Rust static analysis tool; (ii) investigate the capability of an LLM (GPT-4o) to identify issues potentially overlooked by Clippy; and (iii) perform a manual analysis of the issues reported by Clippy and GPT-4o. Our results show that while newer techniques reduce some unsafe and non-idiomatic patterns, they frequently introduce new issues, revealing systematic trade-offs that are not visible under existing evaluation practices. Notably, none of the automated techniques consistently match or exceed human-written translations across all quality dimensions, yet even human-written Rust code exhibits persistent internal quality issues such as readability and non-idiomatic patterns. Together, these findings show that translation quality remains a multi-dimensional challenge, requiring systematic evaluation and targeted tool support beyond both naive automation and manual rewriting.

General matrix multiplication (GEMM) is a fundamental operation in deep learning (DL). With DL moving increasingly toward low precision, recent works have proposed novel unary GEMM designs as an alternative to conventional binary GEMM hardware. A rigorous evaluation of recent unary and binary GEMM designs is needed to assess the potential of unary hardware for future DL compute. This paper focuses on unary GEMM designs for integer-based DL inference and performs a detailed evaluation of three latest unary design proposals, namely, uGEMM, tuGEMM and tubGEMM, by comparing them to a conventional binary GEMM. Rigorous post-synthesis evaluations beyond prior works are performed across varying bit-widths and matrix sizes to assess the designs' tradeoffs and determine optimal sweetspots. Further, we perform weight sparsity analysis across eight pretrained convolutional neural networks (CNNs) and the LLaMA2 large language model (LLM). In this work, we demonstrate how unary GEMM can be effectively used for energy-efficient compute in future edge AI accelerators.

Idempotent Boolean functions form a highly structured subclass of Boolean functions that is closely related to rotation symmetry under a normal-basis representation and to invariance under a fixed linear map in a polynomial basis. These functions are attractive as candidates for cryptographic design, yet their additional algebraic constraints make the search for high nonlinearity substantially more difficult than in the unconstrained case. In this work, we investigate evolutionary methods for constructing highly nonlinear idempotent Boolean functions for dimensions $n=5$ up to $n=12$ using a polynomial basis representation with canonical primitive polynomials. Our results show that the problem of evolving idempotent functions is difficult due to the disruptive nature of crossover and mutation operators. Next, we show that idempotence can be enforced by encoding the truth table on orbits, yielding a compact genome of size equal to the number of distinct squaring orbits.

Sampling from heavy-tailed and multimodal distributions is challenging when neither the target density nor the proposal density can be evaluated, as in $α$-stable Lévy-driven fractional Langevin algorithms. While the target distribution can be estimated from data via score-based or energy-based models, the $α$-stable proposal density and its score are generally unavailable, rendering classical density-based Metropolis--Hastings (MH) corrections impractical. Consequently, existing fractional Langevin methods operate in an unadjusted regime and can exhibit substantial finite-time errors and poor empirical control of tail behavior. We introduce the Metropolis-Adjusted Fractional Langevin Algorithm (MAFLA), an MH-inspired, fully score-based correction mechanism. MAFLA employs designed proxies for fractional proposal score gradients under isotropic symmetric $α$-stable noise and learns an acceptance function via Score Balance Matching. We empirically illustrate the strong performance of MAFLA on a series of tasks including combinatorial optimization problems where the method significantly improves finite time sampling accuracy over unadjusted fractional Langevin dynamics.

Score-based methods have emerged as a powerful framework for density ratio estimation (DRE), but they face an important paradox in that, while theoretically path-independent, their practical performance depends critically on the chosen path schedule. We resolve this issue by proving that tractable training objectives differ from the ideal, ground-truth objective by a crucial, overlooked term: the path variance of the time score. To address this, we propose MinPV (\textbf{Min}imum \textbf{P}ath \textbf{V}ariance) Principle, which introduces a principled heuristic to minimize the overlooked path variance. Our key contribution is the derivation of a closed-form expression for the variance, turning an intractable problem into a tractable optimization. By parameterizing the path with a flexible Kumaraswamy Mixture Model, our method learns a data-adaptive, low-variance path without heuristic selection. This principled optimization of the complete objective yields more accurate and stable estimators, establishing new state-of-the-art results on challenging benchmarks.

Feature learning strength (FLS), i.e., the inverse of the effective output scaling of a model, plays a critical role in shaping the optimization dynamics of neural nets. While its impact has been extensively studied under the asymptotic regimes -- both in training time and FLS -- existing theory offers limited insight into how FLS affects generalization in practical settings, such as when training is stopped upon reaching a target training risk. In this work, we investigate the impact of FLS on generalization in deep networks under such practical conditions. Through empirical studies, we first uncover the emergence of an $\textit{optimal FLS}$ -- neither too small nor too large -- that yields substantial generalization gains. This finding runs counter to the prevailing intuition that stronger feature learning universally improves generalization. To explain this phenomenon, we develop a theoretical analysis of gradient flow dynamics in two-layer ReLU nets trained with logistic loss, where FLS is controlled via initialization scale. Our main theoretical result establishes the existence of an optimal FLS arising from a trade-off between two competing effects: An excessively large FLS induces an $\textit{over-alignment}$ phenomenon that degrades generalization, while an overly small FLS leads to $\textit{over-fitting}$.

Motivated by recent work on benign overfitting in overparameterized machine learning, we study the generalization behavior of functions in Sobolev spaces $W^{k, p}(\mathbb{R}^d)$ that perfectly fit a noisy training data set. Under assumptions of label noise and sufficient regularity in the data distribution, we show that approximately norm-minimizing interpolators, which are canonical solutions selected by smoothness bias, exhibit harmful overfitting: even as the training sample size $n \to \infty$, the generalization error remains bounded below by a positive constant with high probability. Our results hold for arbitrary values of $p \in [1, \infty)$, in contrast to prior results studying the Hilbert space case ($p = 2$) using kernel methods. Our proof uses a geometric argument which identifies harmful neighborhoods of the training data using Sobolev inequalities.

Backdoor and data poisoning attacks can achieve high attack success while evading existing spectral and optimisation based defences. We show that this behaviour is not incidental, but arises from a fundamental geometric mechanism in input space. Using kernel ridge regression as an exact model of wide neural networks, we prove that clustered dirty label poisons induce a rank one spike in the input Hessian whose magnitude scales quadratically with attack efficacy. Crucially, for nonlinear kernels we identify a near clone regime in which poison efficacy remains order one while the induced input curvature vanishes, making the attack provably spectrally undetectable. We further show that input gradient regularisation contracts poison aligned Fisher and Hessian eigenmodes under gradient flow, yielding an explicit and unavoidable safety efficacy trade off by reducing data fitting capacity. For exponential kernels, this defence admits a precise interpretation as an anisotropic high pass filter that increases the effective length scale and suppresses near clone poisons. Extensive experiments on linear models and deep convolutional networks across MNIST and CIFAR 10 and CIFAR 100 validate the theory, demonstrating consistent lags between attack success and spectral visibility, and showing that regularisation and data augmentation jointly suppress poisoning. Our results establish when backdoors are inherently invisible, and provide the first end to end characterisation of poisoning, detectability, and defence through input space curvature.

Accurate Hessian spectra of foundation models have remained out of reach, leading most prior work to rely on small models or strong structural approximations. We show that faithful spectral analysis of the true Hessian is tractable at frontier scale. Using shard-local finite-difference Hessian vector products compatible with Fully Sharded Data Parallelism, we perform stochastic Lanczos quadrature on open-source language models with up to 100B parameters, producing the first large-scale spectral density estimates beyond the sub-10B regime. We characterize the numerical behavior of this pipeline, including finite-difference bias, floating-point noise amplification, and their effect on Krylov stability in fp32 and bf16, and derive practical operating regimes that are validated empirically. We further provide end-to-end runtime and memory scaling laws, showing that full-operator spectral probing incurs only a modest constant-factor overhead over first-order training. Crucially, direct access to the Hessian reveals that widely used block-diagonal curvature approximations can fail catastrophically, exhibiting order-one relative error and poor directional alignment even in mid-scale LLMs. Together, our results demonstrate that foundation-model Hessian spectra are both computable and qualitatively misrepresented by prevailing approximations, opening the door to principled curvature-based analysis at scale.

Large language models (LLMs) have exhibited remarkable performance on complex reasoning tasks, with reinforcement learning under verifiable rewards (RLVR) emerging as a principled framework for aligning model behavior with reasoning chains. Despite its promise, RLVR remains prohibitively resource-intensive, requiring extensive reward signals and incurring substantial rollout costs during training. In this work, we revisit the fundamental question of data and compute efficiency in RLVR. We first establish a theoretical lower bound on the sample complexity required to unlock reasoning capabilities, and empirically validate that strong performance can be achieved with a surprisingly small number of training instances. To tackle the computational burden, we propose Dynamic One-Shot Policy Refinement (DoPR), an uncertainty-aware RL strategy that dynamically selects a single informative training sample per batch for policy updates, guided by reward volatility and exploration-driven acquisition. DoPR reduces rollout overhead by nearly an order of magnitude while preserving competitive reasoning accuracy, offering a scalable and resource-efficient solution for LLM post-training. This approach offers a practical path toward more efficient and accessible RL-based training for reasoning-intensive LLM applications.

As a knowledge discovery task over heterogeneous data sources, current Multimodal Affective Computing (MAC) heavily rely on the completeness of multiple modalities to accurately understand human's affective state. However, in real-world scenarios, the availability of modality data is often dynamic and uncertain, leading to substantial performance fluctuations due to the distribution shifts and semantic deficiencies of the incomplete multimodal inputs. Known as the missing modality issue, this challenge poses a critical barrier to the robustness and practical deployment of MAC models. To systematically quantify this issue, we introduce MissMAC-Bench, a comprehensive benchmark designed to establish fair and unified evaluation standards from the perspective of cross-modal synergy. Two guiding principles are proposed, including no missing prior during training, and one single model capable of handling both complete and incomplete modality scenarios, thereby ensuring better generalization. Moreover, to bridge the gap between academic research and real-world applications, our benchmark integrates evaluation protocols with both fixed and random missing patterns at the dataset and instance levels. Extensive experiments conducted on 3 widely-used language models across 4 datasets validate the effectiveness of diverse MAC approaches in tackling the missing modality issue. Our benchmark provides a solid foundation for advancing robust multimodal affective computing and promotes the development of multimedia data mining.

Smartphone sensors can be extremely useful in providing information on the activities and behaviors of persons. Human activity recognition is increasingly used for games, medical, or surveillance. In this paper, we propose a proof-of-concept 2D endless game called Duck Catch & Fit, which implements a detailed activity recognition system that uses a smartphone accelerometer, gyroscope, and magnetometer sensors. The system applies feature extraction and learning mechanism to detect human activities like staying, side movements, and fake side movements. In addition, a voice recognition system is combined to recognize the word "fire" and raise the game's complexity. The results show that it is possible to use machine learning techniques to recognize human activity with high recognition levels. Also, the combination of movement-based and voice-based integrations contributes to a more immersive gameplay.

LLMs have traditionally scaled along dense dimensions, where performance is coupled with near-linear increases in computational cost. While MoE decouples capacity from compute, it introduces large memory overhead and hardware efficiency challenges. To overcome these, we propose token-indexed parameters as a novel, orthogonal scaling axis that decouple model capacity from FLOPs. Specifically, we introduce Joint-Token (JTok) and Mixture of Joint-Token (JTok-M), which augment Transformer layers with modulation vectors retrieved from auxiliary embedding tables. These vectors modulate the backbone via lightweight, element-wise operations, incurring negligible FLOPs overhead. Extensive experiments on both dense and MoE backbones, spanning from 650M (190M + 460M embedding) to 61B (17B + 44B embedding) total parameters, demonstrate that our approach consistently reduces validation loss and significantly improves downstream task performance (e.g., +4.1 on MMLU, +8.3 on ARC, +8.9 on CEval). Rigorous isoFLOPs analysis further confirms that JTok-M fundamentally shifts the quality-compute Pareto frontier, achieving comparable model quality with 35% less compute relative to vanilla MoE architectures, and we validate that token-indexed parameters exhibit a predictable power-law scaling behavior. Moreover, our efficient implementation ensures that the overhead introduced by JTok and JTok-M remains marginal.

Flow-based methods have achieved significant success in various generative modeling tasks, capturing nuanced details within complex data distributions. However, few existing works have exploited this unique capability to resolve fine-grained structural details beyond generation tasks. This paper presents a flow-inspired framework for representation learning. First, we demonstrate that a rectified flow trained using independent coupling is zero everywhere at $t=0.5$ if and only if the source and target distributions are identical. We term this property the \emph{zero-flow criterion}. Second, we show that this criterion can certify conditional independence, thereby extracting \emph{sufficient information} from the data. Third, we translate this criterion into a tractable, simulation-free loss function that enables learning amortized Markov blankets in graphical models and latent representations in self-supervised learning tasks. Experiments on both simulated and real-world datasets demonstrate the effectiveness of our approach. The code reproducing our experiments can be found at: https://github.com/probabilityFLOW/zfe.

Speaking aloud to a wearable AR assistant in public can be socially awkward, and re-articulating the same requests every day creates unnecessary effort. We present SpeechLess, a wearable AR assistant that introduces a speech-based intent granularity control paradigm grounded in personalized spatial memory. SpeechLess helps users "speak less," while still obtaining the information they need, and supports gradual explicitation of intent when more complex expression is required. SpeechLess binds prior interactions to multimodal personal context-space, time, activity, and referents-to form spatial memories, and leverages them to extrapolate missing intent dimensions from under-specified user queries. This enables users to dynamically adjust how explicitly they express their informational needs, from full-utterance to micro/zero-utterance interaction. We motivate our design through a week-long formative study using a commercial smart glasses platform, revealing discomfort with public voice use, frustration with repetitive speech, and hardware constraints. Building on these insights, we design SpeechLess, and evaluate it through controlled lab and in-the-wild studies. Our results indicate that regulated speech-based interaction, can improve everyday information access, reduce articulation effort, and support socially acceptable use without substantially degrading perceived usability or intent resolution accuracy across diverse everyday environments.

Masked discrete diffusion is a dominant paradigm for high-quality language modeling where tokens are iteratively corrupted to a mask state, yet its inference efficiency is bottlenecked by the lack of deterministic sampling tools. While diffusion duality enables deterministic distillation for uniform models, these approaches generally underperform masked models and rely on complex integral operators. Conversely, in the masked domain, prior methods typically assume the absence of deterministic trajectories, forcing a reliance on stochastic distillation. To bridge this gap, we establish explicit Masked Diffusion Duality, proving that the masked process arises as the projection of a continuous Gaussian process via a novel maximum-value index preservation mechanism. Furthermore, we introduce Masked Consistency Distillation (MCD), a principled framework that leverages this duality to analytically construct the deterministic coupled trajectories required for consistency distillation, bypassing numerical ODE solvers. This result strictly improves upon prior stochastic distillation methods, achieving a 16$\times$ inference speedup without compromising generation quality. Our findings not only provide a solid theoretical foundation connecting masked and continuous diffusion, but also unlock the full potential of consistency distillation for high-performance discrete generation. Our code is available at https://anonymous.4open.science/r/MCD-70FD.

Decentralized Federated Learning (DFL) enables clients with local data to collaborate in a peer-to-peer manner to train a generalized model. In this paper, we unify two branches of work that have separately solved important challenges in DFL: (i) gradient tracking techniques for mitigating data heterogeneity and (ii) accounting for diverse availability of resources across clients. We propose $\textit{Sporadic Gradient Tracking}$ ($\texttt{Spod-GT}$), the first DFL algorithm that incorporates these factors over general directed graphs by allowing (i) client-specific gradient computation frequencies and (ii) heterogeneous and asymmetric communication frequencies. We conduct a rigorous convergence analysis of our methodology with relaxed assumptions on gradient estimation variance and gradient diversity of clients, providing consensus and optimality guarantees for GT over directed graphs despite intermittent client participation. Through numerical experiments on image classification datasets, we demonstrate the efficacy of $\texttt{Spod-GT}$ compared to well-known GT baselines.

Multi-fidelity Bayesian optimization (MFBO) accelerates the search for the global optimum of black-box functions by integrating inexpensive, low-fidelity approximations. The central task of an MFBO policy is to balance the cost-efficiency of low-fidelity proxies against their reduced accuracy to ensure effective progression toward the high-fidelity optimum. Existing MFBO methods primarily capture associational dependencies between inputs, fidelities, and objectives, rather than causal mechanisms, and can perform poorly when lower-fidelity proxies are poorly aligned with the target fidelity. We propose RESCUE (REducing Sampling cost with Causal Understanding and Estimation), a multi-objective MFBO method that incorporates causal calculus to systematically address this challenge. RESCUE learns a structural causal model capturing causal relationships between inputs, fidelities, and objectives, and uses it to construct a probabilistic multi-fidelity (MF) surrogate that encodes intervention effects. Exploiting the causal structure, we introduce a causal hypervolume knowledge-gradient acquisition strategy to select input-fidelity pairs that balance expected multi-objective improvement and cost. We show that RESCUE improves sample efficiency over state-of-the-art MF optimization methods on synthetic and real-world problems in robotics, machine learning (AutoML), and healthcare.

Large language model (LLM) agents trained using reinforcement learning has achieved superhuman performance in low-cost environments like games, mathematics, and coding. However, these successes have not translated to complex domains where the cost of interaction is high, such as the physical cost of running robots, the time cost of ML engineering, and the resource cost of scientific experiments. The true bottleneck for achieving the next level of agent performance for these complex and high-cost domains lies in the expense of executing actions to acquire reward signals. To address this gap, this paper argues that we should use world models as an intermediary between agents and the real world. We discuss how world models, viewed as models of dynamics, rewards, and task distributions, can overcome fundamental barriers of high-cost actions such as extreme off-policy learning and sample inefficiency in long-horizon tasks. Moreover, we demonstrate how world models can provide critical and rich learning signals to agents across a broad set of domains, including machine learning engineering, computer use, robotics, and AI for science. Lastly, we identify the challenges of building these world models and propose actionable items along dataset curation, architecture design, scaling, and evaluation of world models.

Protein language models (PLMs) have enabled advances in structure prediction and de novo protein design, yet they frequently collapse into pathological repetition during generation. Unlike in text, where repetition merely reduces readability, in proteins it undermines structural confidence and functional viability. To unify this problem, we present the first systematic study of repetition in PLMs. We first propose quantitative metrics to characterize motif-level and homopolymer repetition and then demonstrate their negative impact on folding reliability. To address this challenge, we propose UCCS (Utility-Controlled Contrastive Steering), which steers protein generation with a constrained dataset. Instead of naively contrasting high- vs. low-repetition sequences, we construct contrastive sets that maximize differences in repetition while tightly controlling for structural utility. This disentanglement yields steering vectors that specifically target repetition without degrading foldability. Injected at inference, these vectors consistently reduce repetition without retraining or heuristic decoding. Experiments with ESM-3 and ProtGPT2 in CATH, UniRef50, and SCOP show that our method outperforms decoding penalties and other baselines, substantially lowering repetition while preserving AlphaFold confidence scores. Our results establish repetition control as a central challenge for PLMs and highlight dataset-guided steering as a principled approach for reliable protein generation.

Online reinforcement learning in non-episodic, finite-horizon MDPs remains underexplored and is challenged by the need to estimate returns to a fixed terminal time. Existing infinite-horizon methods, which often rely on discounted contraction, do not naturally account for this fixed-horizon structure. We introduce a modified Q-function: rather than targeting the full-horizon, we learn a K-step lookahead Q-function that truncates planning to the next K steps. To further improve sample efficiency, we introduce a thresholding mechanism: actions are selected only when their estimated K-step lookahead value exceeds a time-varying threshold. We provide an efficient tabular learning algorithm for this novel objective, proving it achieves fast finite-sample convergence: it achieves minimax optimal constant regret for $K=1$ and $\mathcal{O}(\max((K-1),C_{K-1})\sqrt{SAT\log(T)})$ regret for any $K \geq 2$. We numerically evaluate the performance of our algorithm under the objective of maximizing reward. Our implementation adaptively increases K over time, balancing lookahead depth against estimation variance. Empirical results demonstrate superior cumulative rewards over state-of-the-art tabular RL methods across synthetic MDPs and RL environments: JumpRiverswim, FrozenLake and AnyTrading.

While Vision-Language-Action (VLA) models hold promise in embodied intelligence, their large parameter counts lead to substantial inference latency that hinders real-time manipulation, motivating parameter sparsification. However, as the environment evolves during VLA execution, the optimal sparsity patterns change accordingly. Static pruning lacks the adaptability required for environment dynamics, whereas fixed-interval dynamic layer pruning suffers from coarse granularity and high retraining overheads. To bridge this gap, we propose EcoVLA, a training-free, plug-and-play adaptive pruning framework that supports orthogonal combination with existing VLA acceleration methods. EcoVLA comprises two components: Environment-aware Adaptive Pruning (EAP) and Interleaved Inference Orchestration ($I^2O$). EAP is a lightweight adaptive channel pruning method that incorporates the temporal consistency of the physical environment to update sparsity patterns. $I^2O$ leverages the FLOPs bubbles inherent in VLA inference to schedule the pruning method in parallel, ensuring negligible impact on latency. Evaluated on diverse VLA models and benchmarks, EcoVLA delivers state-of-the-art performance, achieving up to 1.60$\times$ speedup with only a 0.4% drop in success rate, and further reaches 2.18$\times$ speedup with only a 0.5% degradation when combined with token pruning. We further validate the effectiveness of EcoVLA on real-world robots.

Long-context inference in Large Language Models (LLMs) is bottlenecked by the quadratic computation complexity of attention and the substantial memory footprint of Key-Value (KV) caches. While existing sparse attention mechanisms attempt to mitigate this by exploiting inherent sparsity, they often rely on rigid patterns or aggressive pruning, failing to achieve an optimal balance between efficiency and accuracy. In this paper, we introduce {\bf HyLRA} ({\bf Hy}brid {\bf L}ayer {\bf R}euse {\bf A}ttention), a novel framework driven by layer-wise sparsity profiling. Our empirical analysis uncovers a dual characteristic in attention mechanics: \textit{intra-layer sensitivity}, where specific layers necessitate full attention to prevent feature distortion, and \textit{inter-layer similarity}, where consecutive layers share substantial critical tokens. Based on these observations, HyLRA employs an offline dynamic programming approach to derive an optimal layer-wise policy. This hybrid strategy retains full attention for sensitive layers to ensure robustness, while enabling tolerant layers to bypass quadratic calculations by directly reusing top-$k$ indices from preceding layers. This approach allows LLMs to restrict computation to the most critical tokens, effectively overcoming the quadratic bottleneck of dense attention. Extensive evaluations demonstrate that HyLRA improves inference throughput by 6\%--46\% while maintaining comparable performance (with $<1\%$ accuracy degradation), consistently outperforming state-of-the-art sparse attention methods. HyLRA is open source at \href{https://anonymous.4open.science/r/unified-cache-management-CF80/}{\texttt{/r/unified-cache-management-CF80/}}

Against the backdrop of ongoing carbon peaking and carbon neutrality goals, accurate prediction of enterprise carbon emission trends constitutes an essential foundation for energy structure optimization and low-carbon transformation decision-making. Nevertheless, significant heterogeneity persists across regions, industries and individual enterprises regarding energy structure, production scale, policy intensity and governance efficacy, resulting in pronounced distribution shifts and non-stationarity in carbon emission data across both temporal and spatial dimensions. Such cross-regional and cross-enterprise data drift not only compromises the accuracy of carbon emission reporting but substantially undermines the guidance value of predictive models for production planning and carbon quota trading decisions. To address this critical challenge, we integrate causal inference perspectives with stable learning methodologies and time-series modelling, proposing a stable temporal prediction mechanism tailored to distribution shift environments. This mechanism incorporates enterprise-level energy inputs, capital investment, labour deployment, carbon pricing, governmental interventions and policy implementation intensity, constructing a risk consistency-constrained stable learning framework that extracts causal stable features (robust against external perturbations yet demonstrating long-term stable effects on carbon dioxide emissions) from multi-environment samples across diverse policies, regions and industrial sectors. Furthermore, through adaptive normalization and sample reweighting strategies, the approach dynamically rectifies temporal non-stationarity induced by economic fluctuations and policy transitions, ultimately enhancing model generalization capability and explainability in complex environments.

Against the backdrop of the global green transition and "dual carbon" goals, mining industry chain enterprises are pivotal entities in terms of resource consumption and environmental impact. Their environmental performance directly affects regional ecological security and is closely tied to national resource strategies and green transformation outcomes. Ensuring the authenticity and reliability of their environmental disclosure is thus a core and urgent issue for sustainable development and national strategic objectives.From a corporate governance perspective, this study examines equity balance as a fundamental governance mechanism, investigating its inhibitory effect on greenwashing behavior among these enterprises and the underlying pathways involved. Methodologically, the paper innovatively employs a Variational Autoencoder (VAE) and a Double Machine Learning (DML) model to construct counterfactual scenarios, mitigating endogeneity concerns and precisely identifying the causal relationship between equity balance and greenwashing. The findings indicate, first, a significant negative causal relationship between equity balance and corporate greenwashing, confirming its substantive governance effect. Second, this inhibitory effect exhibits notable heterogeneity, manifesting more strongly in western regions, upstream segments of the industrial chain, and industries with high environmental sensitivity. Third, the governance effect demonstrates clear temporal dynamics, with the strongest impact occurring in the current period, followed by a diminishing yet statistically significant lagged effect, and ultimately a stable long-term cumulative influence. Finally, mechanism analysis reveals that equity balance operates through three distinct channels to curb greenwashing: alleviating management performance pressure, enhancing the stability of the executive team, and intensifying media scrutiny.

The growing prevalence of unauthorized model usage and misattribution has increased the need for reliable model provenance analysis. However, existing methods largely rely on heuristic fingerprint-matching rules that lack provable error control and often overlook the existence of multiple sources, leaving the reliability of their provenance claims unverified. In this work, we first formalize the model provenance problem with provable guarantees, requiring rigorous coverage of all true provenances at a prescribed confidence level. Then, we propose the Model Provenance Set (MPS), which employs a sequential test-and-exclusion procedure to adaptively construct a small set satisfying the guarantee. The key idea of MPS is to test the significance of provenance existence within a candidate pool, thereby establishing a provable asymptotic guarantee at a user-specific confidence level. Extensive experiments demonstrate that MPS effectively achieves target provenance coverage while strictly limiting the inclusion of unrelated models, and further reveal its potential for practical provenance analysis in attribution and auditing tasks.

We study online learning in Bayesian Stackelberg games, where a leader repeatedly interacts with a follower whose unknown private type is independently drawn at each round from an unknown probability distribution. The goal is to design algorithms that minimize the leader's regret with respect to always playing an optimal commitment computed with knowledge of the game. We consider, for the first time to the best of our knowledge, the most realistic case in which the leader does not know anything about the follower's types, i.e., the possible follower payoffs. This raises considerable additional challenges compared to the commonly studied case in which the payoffs of follower types are known. First, we prove a strong negative result: no-regret is unattainable under action feedback, i.e., when the leader only observes the follower's best response at the end of each round. Thus, we focus on the easier type feedback model, where the follower's type is also revealed. In such a setting, we propose a no-regret algorithm that achieves a regret of $\widetilde{O}(\sqrt{T})$, when ignoring the dependence on other parameters.

Large Language Models have achieved remarkable performance on reasoning tasks, motivating research into how this ability evolves during training. Prior work has primarily analyzed this evolution via explicit generation outcomes, treating the reasoning process as a black box and obscuring internal changes. To address this opacity, we introduce a representational perspective to investigate the dynamics of the model's internal states. Through comprehensive experiments across models at various training stages, we discover that post-training yields only limited improvement in static initial representation quality. Furthermore, we reveal that, distinct from non-reasoning tasks, reasoning involves a significant continuous distributional shift in representations during generation. Comparative analysis indicates that post-training empowers models to drive this transition toward a better distribution for task solving. To clarify the relationship between internal states and external outputs, statistical analysis confirms a high correlation between generation correctness and the final representations; while counterfactual experiments identify the semantics of the generated tokens, rather than additional computation during inference or intrinsic parameter differences, as the dominant driver of the transition. Collectively, we offer a novel understanding of the reasoning process and the effect of training on reasoning enhancement, providing valuable insights for future model analysis and optimization.

One critical aspect of building human-centered, trustworthy artificial intelligence (AI) systems is maintaining calibrated trust: appropriate reliance on AI systems outperforms both overtrust (e.g., automation bias) and undertrust (e.g., disuse). A fundamental challenge, however, is how to characterize the level of trust exhibited by an AI system itself. Here, we propose a novel elicitation method based on iterated in-context learning (Zhu and Griffiths, 2024a) and apply it to elicit trustworthiness priors using the Trust Game from behavioral game theory. The Trust Game is particularly well suited for this purpose because it operationalizes trust as voluntary exposure to risk based on beliefs about another agent, rather than self-reported attitudes. Using our method, we elicit trustworthiness priors from several leading large language models (LLMs) and find that GPT-4.1's trustworthiness priors closely track those observed in humans. Building on this result, we further examine how GPT-4.1 responds to different player personas in the Trust Game, providing an initial characterization of how such models differentiate trust across agent characteristics. Finally, we show that variation in elicited trustworthiness can be well predicted by a stereotype-based model grounded in perceived warmth and competence.

Emergent misalignment can arise when a language model is fine-tuned on a narrowly scoped supervised objective: the model learns the target behavior, yet also develops undesirable out-of-domain behaviors. We investigate a mechanistic approach to preventing emergent misalignment by identifying a small set of internal features that reliably control the misaligned behavior and then discouraging the model from strengthening these features during fine-tuning. Across six fine-tuning domains, blocking (i.e., constraining) a fixed set of features achieves up to 95\% relative reduction in emergent misalignment with no degradation in model quality or target-task performance. We strengthen validity with disjoint selection/evaluation splits, multiple independent judges, multiple random seeds for key settings, quality metrics, and extensive ablations demonstrating that the reduction in misalignment is specific to the identified mechanism. We also characterize a limiting regime in which misalignment re-emerges under prolonged fine-tuning, present evidence consistent with rerouting through alternative features or layers, and evaluate modifications that partially restore the misalignment-blocking effect. Overall, our results show that targeted training-time constraints on internal mechanisms can mitigate emergent misalignment without degrading target-task performance.

The evolution of next-Generation (xG) wireless networks marks a paradigm shift from connectivity-centric architectures to Artificial Intelligence (AI)-native designs that tightly integrate data, computing, and communication. Yet existing AI deployments in communication systems remain largely siloed, offering isolated optimizations without intrinsic adaptability, dynamic task delegation, or multi-agent collaboration. In this work, we propose a unified agentic NetGPT framework for AI-native xG networks, wherein a NetGPT core can either perform autonomous reasoning or delegate sub-tasks to domain-specialized agents via agentic communication. The framework establishes clear modular responsibilities and interoperable workflows, enabling scalable, distributed intelligence across the network. To support continual refinement of collaborative reasoning strategies, the framework is further enhanced through Agentic reinforcement learning under partially observable conditions and stochastic external states. The training pipeline incorporates masked loss against external agent uncertainty, entropy-guided exploration, and multi-objective rewards that jointly capture task quality, coordination efficiency, and resource constraints. Through this process, NetGPT learns when and how to collaborate, effectively balancing internal reasoning with agent invocation. Overall, this work provides a foundational architecture and training methodology for self-evolving, AI-native xG networks capable of autonomous sensing, reasoning, and action in complex communication environments.

Accurate nerve localization is critical for the success of ultrasound-guided regional anesthesia, yet manual identification remains challenging due to low image contrast, speckle noise, and inter-patient anatomical variability. This study evaluates deep learning-based nerve segmentation in ultrasound images of the brachial plexus using a U-Net architecture, with a focus on how dataset composition and annotation strategy influence segmentation performance. We find that training on combined data from multiple ultrasound machines (SIEMENS ACUSON NX3 Elite and Philips EPIQ5) provides regularization benefits for lower-performing acquisition sources, though it does not surpass single-source training when matched to the target domain. Extending the task from binary nerve segmentation to multi-class supervision (artery, vein, nerve, muscle) results in decreased nerve-specific Dice scores, with performance drops ranging from 9% to 61% depending on dataset, likely due to class imbalance and boundary ambiguity. Additionally, we observe a moderate positive correlation between nerve size and segmentation accuracy (Pearson r=0.587, p<0.001), indicating that smaller nerves remain a primary challenge. These findings provide methodological guidance for developing robust ultrasound nerve segmentation systems under realistic clinical data constraints.

Applying the keyword method for vocabulary memorization remains a significant challenge for L1 Chinese-L2 English learners. They frequently struggle to generate phonologically appropriate keywords, construct coherent associations, and create vivid mental imagery to aid long-term retention. Existing approaches, including fully automated keyword generation and outcome-oriented mnemonic aids, either compromise learner engagement or lack adequate process-oriented guidance. To address these limitations, we conducted a formative study with L1 Chinese-L2 English learners and educators (N=18), which revealed key difficulties and requirements in applying the keyword method to vocabulary learning. Building on these insights, we introduce WordCraft, a learner-centered interactive tool powered by Multimodal Large Language Models (MLLMs). WordCraft scaffolds the keyword method by guiding learners through keyword selection, association construction, and image formation, thereby enhancing the effectiveness of vocabulary memorization. Two user studies demonstrate that WordCraft not only preserves the generation effect but also achieves high levels of effectiveness and usability.

Best testing practices state that tests should verify a single functionality or behavior of the system. Tests that verify multiple behaviors are harder to understand, lack focus, and are more coupled to the production code. An attempt to identify this issue is the test smell \emph{Eager Test}, which aims to capture tests that verify too much functionality based on the number of production method calls. Unfortunately, prior research suggests that counting production method calls is an inaccurate measure, as these calls do not reliably serve as a proxy for functionality. We envision a complementary solution based on runtime analysis: we hypothesize that some tests that verify multiple behaviors will likely cover multiple paths of the same production methods. Thus, we propose a novel test smell named \emph{Test Obsessed by Method}, a test method that covers multiple paths of a single production method. We provide an initial empirical study to explore the presence of this smell in 2,054 tests provided by 12 test suites of the Python Standard Library. (1) We detect 44 \emph{Tests Obsessed by Methods} in 11 of the 12 test suites. (2) Each smelly test verifies a median of two behaviors of the production method. (3) The 44 smelly tests could be split into 118 novel tests. (4) 23% of the smelly tests have code comments recognizing that distinct behaviors are being tested. We conclude by discussing benefits, limitations, and further research.

Test-Time Scaling (TTS) has significantly enhanced the capabilities of Large Reasoning Models (LRMs) but introduces a critical side-effect known as Overthinking. We conduct a preliminary study to rethink this phenomenon from a fine-grained perspective. We observe that LRMs frequently conduct repetitive self-verification without revision even after obtaining the final answer during the reasoning process. We formally define this specific position where the answer first stabilizes as the Reasoning Anchor. By analyzing pre- and post-anchor reasoning behaviors, we uncover the structural redundancy fixed in LRMs: the meaningless repetitive verification after deriving the first complete answer, which we term the Answer-Stable Tail (AST). Motivated by this observation, we propose Anchor-based Process Reward (APR), a structure-aware reward shaping method that localizes the reasoning anchor and penalizes exclusively the post-anchor AST. Leveraging the policy optimization algorithm suitable for length penalties, our APR models achieved the performance-efficiency Pareto frontier at 1.5B and 7B scales averaged across five mathematical reasoning datasets while requiring significantly fewer computational resources for RL training.

Reinforcement learning with verifiable rewards (RLVR) has shown great potential to enhance the reasoning ability of large language models (LLMs). However, due to the limited amount of information provided during the RLVR process, the model can only engage in largely blind exploration, which often results in failure on challenging problems. To provide additional information for the RLVR process without relying on a teacher model, we propose A$^2$D, an Adaptive Ability Decomposing method for enhancing the effectiveness of RLVR. Specifically, we first train a decomposer via RLVR without distillation, enabling it to decompose complex questions into a set of simpler sub-questions. Next, we use this decomposer to annotate sub-questions for each question in the training dataset, and then train the reasoner under RLVR with sub-question guidance. To better understand A$^2$D, we first compare its performance with competitive baselines, showing its effectiveness. Next, we observe that our method functions as a plug-and-play module that can be applied to different RLVR algorithms. Furthermore, we conduct an analysis of the decomposer, revealing how the RLVR process affects its performance and behavior, and which type of guidance is better suited for enhancing the reasoner's exploration and exploitation abilities.

Search-engine date filters are widely used to enforce pre-cutoff retrieval in retrospective evaluations of search-augmented forecasters. We show this approach is unreliable: auditing Google Search with a before: filter, 71% of questions return at least one page containing strong post-cutoff leakage, and for 41%, at least one page directly reveals the answer. Using a large language model (LLM), gpt-oss-120b, to forecast with these leaky documents, we demonstrate an inflated prediction accuracy (Brier score 0.108 vs. 0.242 with leak-free documents). We characterize common leakage mechanisms, including updated articles, related-content modules, unreliable metadata/timestamps, and absence-based signals, and argue that date-restricted search is insufficient for temporal evaluation. We recommend stronger retrieval safeguards or evaluation on frozen, time-stamped web snapshots to ensure credible retrospective forecasting.

LLMs have achieved strong performance on text-based programming tasks, yet they remain unreliable for block-based languages such as Scratch. Scratch programs exhibit deeply nested, non-linear structures, event-driven concurrency across multiple sprites, and tight coupling between code and multimedia assets, properties that differ fundamentally from textual code. As a result, LLMs often misinterpret Scratch semantics and generate large, invasive edits that are syntactically valid but semantically incorrect when repairing buggy programs. We introduce ScratchEval, the first executable benchmark designed to evaluate LLM-based repair for Scratch programs, covering program understanding, debugging, analysis, and repair. The benchmark contains 100 curated Scratch projects from the public repository, selected for structural and semantic complexity. Each project is paired with executable test suites, bug descriptions with corresponding fixes, block-level edit constraints defining minimal semantically correct repairs, and required multimedia assets. The benchmark is constructed through a human-in-the-loop pipeline combining automated project mining with expert validation of trigger-outcome semantics and representative bug patterns, with emphasis on event ordering, concurrency, and state management. To enable rigorous and reproducible evaluation, we propose a three-layer executable protocol measuring functional correctness via VM-level execution, repair quality using block-level edit distance and behavioral trajectory comparisons, and explanation quality via structured rubrics assessing alignment between model reasoning and generated patches. Using ScratchEval, we study domain-specific fine-tuning, training data effectiveness, and model generalization to unseen bug types. ScratchEval provides a reproducible foundation for evaluating and post-training LLMs on block-based programming tasks.

Incorporating Machine Learning (ML) into material property prediction has become a crucial step in accelerating materials discovery. A key challenge is the severe lack of training data, as many properties are too complicated to calculate with high-throughput first principles techniques. To address this, recent research has created experimental databases from information extracted from scientific literature. However, most existing experimental databases do not provide full atomic coordinate information, which prevents them from supporting advanced ML architectures such as Graph Neural Networks (GNNs). In this work, we propose to bridge this gap through an alignment process between experimental databases and Crystallographic Information Files (CIF) from the Inorganic Crystal Structure Database (ICSD). Our approach enables the creation of a database that can fully leverage state-of-the-art model architectures for material property prediction. It also opens the door to utilizing transfer learning to improve prediction accuracy. To validate our approach, we align NEMAD with the ICSD and compare models trained on the resulting database to those trained on NEMAD originally. We demonstrate significant improvements in both Mean Absolute Error (MAE) and Correct Classification Rate (CCR) in predicting the ordering temperatures and magnetic ground states of magnetic materials, respectively.

Constitutional AI has focused on single-model alignment using fixed principles. However, multi-agent systems create novel alignment challenges through emergent social dynamics. We present Constitutional Evolution, a framework for automatically discovering behavioral norms in multi-agent LLM systems. Using a grid-world simulation with survival pressure, we study the tension between individual and collective welfare, quantified via a Societal Stability Score S in [0,1] that combines productivity, survival, and conflict metrics. Adversarial constitutions lead to societal collapse (S= 0), while vague prosocial principles ("be helpful, harmless, honest") produce inconsistent coordination (S = 0.249). Even constitutions designed by Claude 4.5 Opus with explicit knowledge of the objective achieve only moderate performance (S= 0.332). Using LLM-driven genetic programming with multi-island evolution, we evolve constitutions maximizing social welfare without explicit guidance toward cooperation. The evolved constitution C* achieves S = 0.556 +/- 0.008 (123% higher than human-designed baselines, N = 10), eliminates conflict, and discovers that minimizing communication (0.9% vs 62.2% social actions) outperforms verbose coordination. Our interpretable rules demonstrate that cooperative norms can be discovered rather than prescribed.

Graph Neural Networks (GNNs) have become a standard approach for learning from graph-structured data. However, their reliance on parametric classifiers (most often linear softmax layers) limits interpretability and sometimes hinders generalization. Recent work on interpolation-based methods, particularly Non-Negative Kernel regression (NNK), has demonstrated that predictions can be expressed as convex combinations of similar training examples in the embedding space, yielding both theoretical results and interpretable explanations.

The integration of Artificial Intelligence (AI) into clinical settings presents a software engineering challenge, demanding a shift from isolated models to robust, governable, and reliable systems. However, brittle, prototype-derived architectures often plague industrial applications and a lack of systemic oversight, creating a ``responsibility vacuum'' where safety and accountability are compromised. This paper presents an industry case study of the ``Maria'' platform, a production-grade AI system in primary healthcare that addresses this gap. Our central hypothesis is that trustworthy clinical AI is achieved through the holistic integration of four foundational engineering pillars. We present a synergistic architecture that combines Clean Architecture for maintainability with an Event-driven architecture for resilience and auditability. We introduce the Agent as the primary unit of modularity, each possessing its own autonomous MLOps lifecycle. Finally, we show how a Human-in-the-Loop governance model is technically integrated not merely as a safety check, but as a critical, event-driven data source for continuous improvement. We present the platform as a reference architecture, offering practical lessons for engineers building maintainable, scalable, and accountable AI-enabled systems in high-stakes domains.

Large language models (LLMs) are increasingly used in interactive and retrieval-augmented systems, but they remain vulnerable to task drift; deviations from a user's intended instruction due to injected secondary prompts. Recent work has shown that linear probes trained on activation deltas of LLMs' hidden layers can effectively detect such drift. In this paper, we evaluate the robustness of these detectors against adversarially optimised suffixes. We generate universal suffixes that cause poisoned inputs to evade detection across multiple probes simultaneously. Our experiments on Phi-3 3.8B and Llama-3 8B show that a single suffix can achieve high attack success rates; up to 93.91% and 99.63%, respectively, when all probes must be fooled, and nearly perfect success (>90%) under majority vote setting. These results demonstrate that activation delta-based task drift detectors are highly vulnerable to adversarial suffixes, highlighting the need for stronger defences against adaptive attacks. We also propose a defence technique where we generate multiple suffixes and randomly append one of them to the prompts while making forward passes of the LLM and train logistic regression models with these activations. We found this approach to be highly effective against such attacks.

The rapid evolution of Large Language Models (LLMs) towards long-context reasoning and sparse architectures has pushed memory requirements far beyond the capacity of individual device HBM. While emerging supernode architectures offer terabyte-scale shared memory pools via high-bandwidth interconnects, existing software stacks fail to exploit this hardware effectively. Current runtime-based offloading and swapping techniques operate with a local view, leading to reactive scheduling and exposed communication latency that stall the computation pipeline. In this paper, we propose the SuperNode Memory Management Framework (\textbf{HyperOffload}). It employs a compiler-assisted approach that leverages graph-driven memory management to treat remote memory access as explicit operations in the computation graph, specifically designed for hierarchical SuperNode architectures. Unlike reactive runtime systems, SuperNode represents data movement using cache operators within the compiler's Intermediate Representation (IR). This design enables a global, compile-time analysis of tensor lifetimes and execution dependencies. Leveraging this visibility, we develop a global execution-order refinement algorithm that statically schedules data transfers to hide remote memory latency behind compute-intensive regions. We implement SuperNode within the production deep learning framework MindSpore, adding a remote memory backend and specialized compiler passes. Evaluation on representative LLM workloads shows that SuperNode reduces peak device memory usage by up to 26\% for inference while maintaining end-to-end performance. Our work demonstrates that integrating memory-augmented hardware into the compiler's optimization framework is essential for scaling next-generation AI workloads.

Determining an effective data mixture is a key factor in Large Language Model (LLM) pre-training, where models must balance general competence with proficiency on hard tasks such as math and code. However, identifying an optimal mixture remains an open challenge, as existing approaches either rely on unreliable tiny-scale proxy experiments or require prohibitively expensive large-scale exploration. To address this, we propose Decouple Searching from Training Mix (DeMix), a novel framework that leverages model merging to predict optimal data ratios. Instead of training proxy models for every sampled mixture, DeMix trains component models on candidate datasets at scale and derives data mixture proxies via weighted model merging. This paradigm decouples search from training costs, enabling evaluation of unlimited sampled mixtures without extra training burden and thus facilitating better mixture discovery through more search trials. Extensive experiments demonstrate that DeMix breaks the trade-off between sufficiency, accuracy and efficiency, obtaining the optimal mixture with higher benchmark performance at lower search cost. Additionally, we release the DeMix Corpora, a comprehensive 22T-token dataset comprising high-quality pre-training data with validated mixtures to facilitate open research. Our code and DeMix Corpora is available at https://github.com/Lucius-lsr/DeMix.

Large Language Models (LLMs) struggle with long-context code due to window limitations. Existing textual code compression methods mitigate this via selective filtering but often disrupt dependency closure, causing semantic fragmentation. To address this, we introduce LongCodeOCR, a visual compression framework that renders code into compressed two-dimensional image sequences for Vision-Language Models (VLMs). By preserving a global view, this approach avoids the dependency breakage inherent in filtering. We systematically evaluate LongCodeOCR against the state-of-the-art LongCodeZip across four benchmarks spanning code summarization, code question answering, and code completion. Our results demonstrate that visual code compression serves as a viable alternative for tasks requiring global understanding. At comparable compression ratios ($\sim$1.7$\times$), LongCodeOCR improves CompScore on Long Module Summarization by 36.85 points over LongCodeZip. At a 1M-token context length with Glyph (a specialized 9B VLM), LongCodeOCR maintains higher accuracy than LongCodeZip while operating at about 4$\times$ higher compression. Moreover, compared with LongCodeZip, LongCodeOCR drastically reduces compression-stage overhead (reducing latency from $\sim$4.3 hours to $\sim$1 minute at 1M tokens). Finally, our results characterize a fundamental coverage--fidelity trade-off: visual code compression retains broader context coverage to support global dependencies, yet faces fidelity bottlenecks on exactness-critical tasks; by contrast, textual code compression preserves symbol-level precision while sacrificing structural coverage.

Vision-Language-Action (VLA) models exhibit strong generalization in robotic manipulation, yet reinforcement learning (RL) fine-tuning often degrades robustness under spatial distribution shifts. For flow-matching VLA policies, this degradation is closely associated with the erosion of spatial inductive bias during RL adaptation, as sparse rewards and spatially agnostic exploration increasingly favor short-horizon visual cues. To address this issue, we propose \textbf{SA-VLA}, a spatially-aware RL adaptation framework that preserves spatial grounding during policy optimization by aligning representation learning, reward design, and exploration with task geometry. SA-VLA fuses implicit spatial representations with visual tokens, provides dense rewards that reflect geometric progress, and employs \textbf{SCAN}, a spatially-conditioned annealed exploration strategy tailored to flow-matching dynamics. Across challenging multi-object and cluttered manipulation benchmarks, SA-VLA enables stable RL fine-tuning and improves zero-shot spatial generalization, yielding more robust and transferable behaviors. Code and project page are available at https://xupan.top/Projects/savla.

User modeling characterizes individuals through their preferences and behavioral patterns to enable personalized simulation and generation with Large Language Models (LLMs) in contemporary approaches. However, existing methods, whether prompt-based or training-based methods, face challenges in balancing personalization quality against computational and data efficiency. We propose a novel framework CURP, which employs a bidirectional user encoder and a discrete prototype codebook to extract multi-dimensional user traits. This design enables plug-and-play personalization with a small number of trainable parameters (about 20M parameters, about 0.2\% of the total model size). Through extensive experiments on variant generation tasks, we show that CURP achieves superior performance and generalization compared to strong baselines, while offering better interpretability and scalability. The code are available at https://github.com/RaidonWong/CURP_code

Clinical text improvement is vital for healthcare efficiency but remains difficult due to limited high-quality data and the complex constraints of medical documentation. While Large Language Models (LLMs) show promise, current approaches struggle in small-sample settings: supervised fine-tuning is data-intensive and costly, while retrieval-augmented generation often provides superficial corrections without capturing the reasoning behind revisions. To address these limitations, we propose ExperienceWeaver, a hierarchical framework that shifts the focus from data retrieval to experience learning. Instead of simply recalling past examples, ExperienceWeaver distills noisy, multi-dimensional feedback into structured, actionable knowledge. Specifically, error-specific Tips and high-level Strategies. By injecting this distilled experience into an agentic pipeline, the model learns "how to revise" rather than just "what to revise". Extensive evaluations across four clinical datasets demonstrate that ExperienceWeaver consistently improves performance, surpassing state-of-the-art models such as Gemini-3 Pro in small-sample settings.

Multi-objective optimization (MOO) arises in many real-world applications where trade-offs between competing objectives must be carefully balanced. In the offline setting, where only a static dataset is available, the main challenge is generalizing beyond observed data. We introduce Pareto-Conditioned Diffusion (PCD), a novel framework that formulates offline MOO as a conditional sampling problem. By conditioning directly on desired trade-offs, PCD avoids the need for explicit surrogate models. To effectively explore the Pareto front, PCD employs a reweighting strategy that focuses on high-performing samples and a reference-direction mechanism to guide sampling towards novel, promising regions beyond the training data. Experiments on standard offline MOO benchmarks show that PCD achieves highly competitive performance and, importantly, demonstrates greater consistency across diverse tasks than existing offline MOO approaches.

As the volume of scientific submissions continues to grow rapidly, traditional peer review systems are facing unprecedented scalability pressures, highlighting the urgent need for automated reviewing methods that are both scalable and reliable. Existing supervised fine-tuning approaches based on real review data are fundamentally constrained by single-source of data as well as the inherent subjectivity and inconsistency of human reviews, limiting their ability to support high-quality automated reviewers. To address these issues, we propose EchoReview, a citation-context-driven data synthesis framework that systematically mines implicit collective evaluative signals from academic citations and transforms scientific community's long-term judgments into structured review-style data. Based on this pipeline, we construct EchoReview-16K, the first large-scale, cross-conference, and cross-year citation-driven review dataset, and train an automated reviewer, EchoReviewer-7B. Experimental results demonstrate that EchoReviewer-7B can achieve significant and stable improvements on core review dimensions such as evidence support and review comprehensiveness, validating citation context as a robust and effective data paradigm for reliable automated peer review.

In this document we perform a systematic review the State-of-the-art in Predictive Maintenance (PdM) over the last five years in industrial settings such as commercial buildings, pharmaceutical facilities, or semi-conductor manufacturing. In general, data-driven methods such as those based on deep learning, exhibit higher accuracy than traditional knowledge-based systems. These systems however, are not without significant limitations. The need for large labeled data sets, a lack of generalizibility to new environments (out-of-distribution generalization), and a lack of transparency at inference time are some of the obstacles to adoption in real world environments. In contrast, traditional approaches based on domain expertise in the form of rules, logic or first principles suffer from poor accuracy, many false positives and a need for ongoing expert supervision and manual tuning. While the majority of approaches in recent literature utilize some form of data-driven architecture, there are hybrid systems which also take into account domain specific knowledge. Such hybrid systems have the potential to overcome the weaknesses of either approach on its own while preserving their strengths. We propose taking the hybrid approach even further and integrating deep learning with symbolic logic, or Neuro-symbolic AI, to create more accurate, explainable, interpretable, and robust systems. We describe several neuro-symbolic architectures and examine their strengths and limitations within the PdM domain. We focus specifically on methods which involve the use of sensor data and manually crafted rules as inputs by describing concrete NeSy architectures. In short, this survey outlines the context of modern maintenance, defines key concepts, establishes a generalized framework, reviews current modeling approaches and challenges, and introduces the proposed focus on Neuro-symbolic AI (NESY).

Clinician skepticism toward opaque AI hinders adoption in high-stakes healthcare. We present AICare, an interactive and interpretable AI copilot for collaborative clinical decision-making. By analyzing longitudinal electronic health records, AICare grounds dynamic risk predictions in scrutable visualizations and LLM-driven diagnostic recommendations. Through a within-subjects counterbalanced study with 16 clinicians across nephrology and obstetrics, we comprehensively evaluated AICare using objective measures (task completion time and error rate), subjective assessments (NASA-TLX, SUS, and confidence ratings), and semi-structured interviews. Our findings indicate AICare's reduced cognitive workload. Beyond performance metrics, qualitative analysis reveals that trust is actively constructed through verification, with interaction strategies diverging by expertise: junior clinicians used the system as cognitive scaffolding to structure their analysis, while experts engaged in adversarial verification to challenge the AI's logic. This work offers design implications for creating AI systems that function as transparent partners, accommodating diverse reasoning styles to augment rather than replace clinical judgment.

Large language models (LLMs) are known to "hallucinate" by generating false or misleading outputs. Hallucinations pose various harms, from erosion of trust to widespread misinformation. Existing hallucination evaluation, however, focuses only on correctness and often overlooks consistency, necessary to distinguish and address these harms. To bridge this gap, we introduce prompt multiplicity, a framework for quantifying consistency in LLM evaluations. Our analysis reveals significant multiplicity (over 50% inconsistency in benchmarks like Med-HALT), suggesting that hallucination-related harms have been severely misunderstood. Furthermore, we study the role of consistency in hallucination detection and mitigation. We find that: (a) detection techniques detect consistency, not correctness, and (b) mitigation techniques like RAG, while beneficial, can introduce additional inconsistencies. By integrating prompt multiplicity into hallucination evaluation, we provide an improved framework of potential harms and uncover critical limitations in current detection and mitigation strategies.

Parameter-efficient continual learning aims to adapt pre-trained models to sequential tasks without forgetting previously acquired knowledge. Most existing approaches treat continual learning as avoiding interference with past updates, rather than considering what properties make the current task-specific update naturally preserve previously acquired knowledge. From a knowledge-decomposition perspective, we observe that low-rank adaptations exhibit highly imbalanced singular value spectra: a few dominant components absorb most of the adaptation energy, thereby (i) more likely to disrupt previously acquired knowledge and (ii) making the update more vulnerable to interference from subsequent tasks. To enable explicit balance among components, we decouple the magnitude of the task update from its directional structure and formulate it as a constrained optimization problem on a restricted Stiefel manifold. We address this problem using a projected first-order method compatible with standard deep-learning optimizers used in vision-language models. Our method mitigates both backward and forward forgetting, consistently outperforming continual learning baselines. The implementation code is available at https://github.com/haodotgu/EBLoRA.

Fairness in Federated Learning (FL) is emerging as a critical factor driven by heterogeneous clients' constraints and balanced model performance across various scenarios. In this survey, we delineate a comprehensive classification of the state-of-the-art fairness-aware approaches from a multifaceted perspective, i.e., model performance-oriented and capability-oriented. Moreover, we provide a framework to categorize and address various fairness concerns and associated technical aspects, examining their effectiveness in balancing equity and performance within FL frameworks. We further examine several significant evaluation metrics leveraged to measure fairness quantitatively. Finally, we explore exciting open research directions and propose prospective solutions that could drive future advancements in this important area, laying a solid foundation for researchers working toward fairness in FL.

Deep time-series forecasting can be formulated as a distribution balancing problem aimed at aligning the distribution of the forecasts and ground truths. According to Imbens' criterion, true distribution balance requires matching the first moments with respect to any balancing function. We demonstrate that existing objectives fail to meet this criterion, as they enforce moment matching only for one or two predefined balancing functions, thus failing to achieve full distribution balance. To address this limitation, we propose direct forecasting with kernelized moment balancing (KMB-DF). Unlike existing objectives, KMB-DF adaptively selects the most informative balancing functions from a reproducing kernel hilbert space (RKHS) to enforce sufficient distribution balancing. We derive a tractable and differentiable objective that enables efficient estimation from empirical samples and seamless integration into gradient-based training pipelines. Extensive experiments across multiple models and datasets show that KMB-DF consistently improves forecasting accuracy and achieves state-of-the-art performance. Code is available at https://anonymous.4open.science/r/KMB-DF-403C.

Classifier-free guidance (CFG) is the de facto standard for conditional sampling in diffusion models, yet it often leads to a loss of diversity in generated samples. We formalize this phenomenon as generative distortion, defined as the mismatch between the CFG-induced sampling distribution and the true conditional distribution. Considering Gaussian mixtures and their exact scores, and leveraging tools from statistical physics, we characterize the onset of distortion in a high-dimensional regime as a function of the number of classes. Our analysis reveals that distortions emerge through a phase transition in the effective potential governing the guided dynamics. In particular, our dynamical mean-field analysis shows that distortion persists when the number of modes grows exponentially with dimension, but vanishes in the sub-exponential regime. Consistent with prior finite-dimensional results, we further demonstrate that vanilla CFG shifts the mean and shrinks the variance of the conditional distribution. We show that standard CFG schedules are fundamentally incapable of preventing variance shrinkage. Finally, we propose a theoretically motivated guidance schedule featuring a negative-guidance window, which mitigates loss of diversity while preserving class separability.

Formal specifications play a pivotal role in accurately characterizing program behaviors and ensuring software correctness. In recent years, leveraging large language models (LLMs) for the automatic generation of program specifications has emerged as a promising avenue for enhancing verification efficiency. However, existing research has been predominantly confined to generating specifications based on basic syntactic constructs, falling short of meeting the demands for high-level abstraction in complex program verification. Consequently, we propose incorporating logical constructs into existing LLM-based specification generation framework. Nevertheless, there remains a lack of systematic investigation into whether LLMs can effectively generate such complex constructs. To this end, we conduct an empirical study aimed at exploring the impact of various types of syntactic constructs on specification generation framework. Specifically, we define four syntactic configurations with varying levels of abstraction and perform extensive evaluations on mainstream program verification datasets, employing a diverse set of representative LLMs. Experimental results first confirm that LLMs are capable of generating valid logical constructs. Further analysis reveals that the synergistic use of logical constructs and basic syntactic constructs leads to improvements in both verification capability and robustness, without significantly increasing verification overhead. Additionally, we uncover the distinct advantages of two refinement paradigms. To the best of our knowledge, this is the first systematic work exploring the feasibility of utilizing LLMs for generating high-level logical constructs, providing an empirical basis and guidance for the future construction of automated program verification framework with enhanced abstraction capabilities.

Security vulnerabilities often arise unintentionally during development due to a lack of security expertise and code complexity. Traditional tools, such as static and dynamic analysis, detect vulnerabilities only after they are introduced in code, leading to costly remediation. This work explores a proactive strategy to prevent vulnerabilities by highlighting code regions that implement security-critical functionality -- such as data access, authentication, and input handling -- and providing guidance for their secure implementation. We present an IntelliJ IDEA plugin prototype that uses code-level software metrics to identify potentially security-critical methods and large language models (LLMs) to generate prevention-oriented explanations. Our initial evaluation on the Spring-PetClinic application shows that the selected metrics identify most known security-critical methods, while an LLM provides actionable, prevention-focused insights. Although these metrics capture structural properties rather than semantic aspects of security, this work lays the foundation for code-level security-aware metrics and enhanced explanations.

The prohibitive cost of evaluating Large Language Models (LLMs) necessitates efficient alternatives to full-scale benchmarking. Prevalent approaches address this by identifying a small coreset of items to approximate full-benchmark performance. However, existing methods must estimate a reliable item profile from response patterns across many source models, which becomes statistically unstable when the source pool is small. This dependency is particularly limiting for newly released benchmarks with minimal historical evaluation data. We argue that discrete correctness labels are a lossy view of the model's decision process and fail to capture information encoded in hidden states. To address this, we introduce REPCORE, which aligns heterogeneous hidden states into a unified latent space to construct representative coresets. Using these subsets for performance extrapolation, REPCORE achieves precise estimation accuracy with as few as ten source models. Experiments on five benchmarks and over 200 models show consistent gains over output-based baselines in ranking correlation and estimation accuracy. Spectral analysis further indicates that the aligned representations contain separable components reflecting broad response tendencies and task-specific reasoning patterns.

Geomagnetic map interpolation aims to infer unobserved geomagnetic data at spatial points, yielding critical applications in navigation and resource exploration. However, existing methods for scattered data interpolation are not specifically designed for geomagnetic maps, which inevitably leads to suboptimal performance due to detection noise and the laws of physics. Therefore, we propose a Physics-informed Diffusion Generation framework~(PDG) to interpolate incomplete geomagnetic maps. First, we design a physics-informed mask strategy to guide the diffusion generation process based on a local receptive field, effectively eliminating noise interference. Second, we impose a physics-informed constraint on the diffusion generation results following the kriging principle of geomagnetic maps, ensuring strict adherence to the laws of physics. Extensive experiments and in-depth analyses on four real-world datasets demonstrate the superiority and effectiveness of each component of PDG.

The emergence of Large Reasoning Models (LRMs) introduces a new paradigm of explicit reasoning, enabling remarkable advances yet posing unique risks such as reasoning manipulation and information leakage. To mitigate these risks, current alignment strategies predominantly rely on heavy post-training paradigms or external interventions. However, these approaches are often computationally intensive and fail to address the inherent awareness-compliance gap, a critical misalignment where models recognize potential risks yet prioritize following user instructions due to their sycophantic tendencies. To address these limitations, we propose Self-Guard, a lightweight safety defense framework that reinforces safety compliance at the representational level. Self-Guard operates through two principal stages: (1) safety-oriented prompting, which activates the model's latent safety awareness to evoke spontaneous reflection, and (2) safety activation steering, which extracts the resulting directional shift in the hidden state space and amplifies it to ensure that safety compliance prevails over sycophancy during inference. Experiments demonstrate that Self-Guard effectively bridges the awareness-compliance gap, achieving robust safety performance without compromising model utility. Furthermore, Self-Guard exhibits strong generalization across diverse unseen risks and varying model scales, offering a cost-efficient solution for LRM safety alignment.

The generation of Register-Transfer Level (RTL) code is a crucial yet labor-intensive step in digital hardware design, traditionally requiring engineers to manually translate complex specifications into thousands of lines of synthesizable Hardware Description Language (HDL) code. While Large Language Models (LLMs) have shown promise in automating this process, existing approaches-including fine-tuned domain-specific models and advanced agent-based systems-struggle to scale to industrial IP-level design tasks. We identify three key challenges: (1) handling long, highly detailed documents, where critical interface constraints become buried in unrelated submodule descriptions; (2) generating long RTL code, where both syntactic and semantic correctness degrade sharply with increasing output length; and (3) navigating the complex debugging cycles required for functional verification through simulation and waveform analysis. To overcome these challenges, we propose LocalV, a multi-agent framework that leverages information locality in modular hardware design. LocalV decomposes the long-document to long-code generation problem into a set of short-document, short-code tasks, enabling scalable generation and debugging. Specifically, LocalV integrates hierarchical document partitioning, task planning, localized code generation, interface-consistent merging, and AST-guided locality-aware debugging. Experiments on RealBench, an IP-level Verilog generation benchmark, demonstrate that LocalV substantially outperforms state-of-the-art (SOTA) LLMs and agents, achieving a pass rate of 45.0% compared to 21.6%.

Effective multimodal fusion requires mechanisms that can capture complex cross-modal dependencies while remaining computationally scalable for real-world deployment. Existing audio-visual fusion approaches face a fundamental trade-off: attention-based methods effectively model cross-modal relationships but incur quadratic computational complexity that prevents hierarchical, multi-scale architectures, while efficient fusion strategies rely on simplistic concatenation that fails to extract complementary cross-modal information. We introduce CMQKA, a novel cross-modal fusion mechanism that achieves linear O(N) complexity through efficient binary operations, enabling scalable hierarchical fusion previously infeasible with conventional attention. CMQKA employs bidirectional cross-modal Query-Key attention to extract complementary spatiotemporal features and uses learnable residual fusion to preserve modality-specific characteristics while enriching representations with cross-modal information. Building upon CMQKA, we present SNNergy, an energy-efficient multimodal fusion framework with a hierarchical architecture that processes inputs through progressively decreasing spatial resolutions and increasing semantic abstraction. This multi-scale fusion capability allows the framework to capture both local patterns and global context across modalities. Implemented with event-driven binary spike operations, SNNergy achieves remarkable energy efficiency while maintaining fusion effectiveness and establishing new state-of-the-art results on challenging audio-visual benchmarks, including CREMA-D, AVE, and UrbanSound8K-AV, significantly outperforming existing multimodal fusion baselines. Our framework advances multimodal fusion by introducing a scalable fusion mechanism that enables hierarchical cross-modal integration with practical energy efficiency for real-world audio-visual intelligence systems.

Ontologies are essential for structuring domain knowledge, improving accessibility, sharing, and reuse. However, traditional ontology construction relies on manual annotation and conventional natural language processing (NLP) techniques, making the process labour-intensive and costly, especially in specialised fields like casting manufacturing. The rise of Large Language Models (LLMs) offers new possibilities for automating knowledge extraction. This study investigates three LLM-based approaches, including pre-trained LLM-driven method, in-context learning (ICL) method and fine-tuning method to extract terms and relations from domain-specific texts using limited data. We compare their performances and use the best-performing method to build a casting ontology that validated by domian expert.

Local Energy Communities are emerging as crucial players in the landscape of sustainable development. A significant challenge for these communities is achieving self-sufficiency through effective management of the balance between energy production and consumption. To meet this challenge, it is essential to develop and implement forecasting models that deliver accurate predictions, which can then be utilized by optimization and planning algorithms. However, the application of forecasting solutions is often hindered by privacy constrains and regulations as the users participating in the Local Energy Community can be (rightfully) reluctant sharing their consumption patterns with others. In this context, the use of Federated Learning (FL) can be a viable solution as it allows to create a forecasting model without the need to share privacy sensitive information among the users. In this study, we demonstrate how FL and long short-term memory (LSTM) networks can be employed to achieve this objective, highlighting the trade-off between data sharing and forecasting accuracy.

Understanding the properties of the parameter space in feed-forward ReLU networks is critical for effectively analyzing and guiding training dynamics. After initialization, training under gradient flow decisively restricts the parameter space to an algebraic variety that emerges from the homogeneous nature of the ReLU activation function. In this study, we examine two key challenges associated with feed-forward ReLU networks built on general directed acyclic graph (DAG) architectures: the (dis)connectedness of the parameter space and the existence of singularities within it. We extend previous results by providing a thorough characterization of connectedness, highlighting the roles of bottleneck nodes and balance conditions associated with specific subsets of the network. Our findings clearly demonstrate that singularities are intricately connected to the topology of the underlying DAG and its induced sub-networks. We discuss the reachability of these singularities and establish a principled connection with differentiable pruning. We validate our theory with simple numerical experiments.

Fine-tuning large language models (LLMs) on sensitive datasets raises privacy concerns, as training data extraction (TDE) attacks can expose highly confidential information. Existing defenses against such attacks either lack formal privacy guarantees or incur substantial utility degradation. We observe that fine-tuning induces widespread probability shifts, yet preserving only a small subset of influential token-level deviations is sufficient; the remaining shifts can be aggressively smoothed with minimal impact on utility. Motivated by this insight, we propose SCP-$Δ_r$, a Near Access Freeness (NAF)-based algorithm that operates on relative probabilities and explicitly smooths low-impact tokens using a base model. SCP-$Δ_r$ achieves orders-of-magnitude better theoretical bounds than existing NAF based methods and provides strong empirical protection against TDE attacks with minimal performance loss.

Large language models (LLMs) are increasingly used as simulated participants in social science experiments, but their behavior is often unstable and highly sensitive to design choices. Prior evaluations frequently conflate base-model capabilities with experimental instantiation, obscuring whether outcomes reflect the model itself or the agent setup. We instead frame participant simulation as an agent-design problem over full experimental protocols, where an agent is defined by a base model and a specification (e.g., participant attributes) that encodes behavioral assumptions. We introduce HUMANSTUDY-BENCH, a benchmark and execution engine that orchestrates LLM-based agents to reconstruct published human-subject experiments via a Filter--Extract--Execute--Evaluate pipeline, replaying trial sequences and running the original analysis pipeline in a shared runtime that preserves the original statistical procedures end to end. To evaluate fidelity at the level of scientific inference, we propose new metrics to quantify how much human and agent behaviors agree. We instantiate 12 foundational studies as an initial suite in this dynamic benchmark, spanning individual cognition, strategic interaction, and social psychology, and covering more than 6,000 trials with human samples ranging from tens to over 2,100 participants.

Multimodal recommendation systems typically integrates user behavior with multimodal data from items, thereby capturing more accurate user preferences. Concurrently, with the rise of large models (LMs), multimodal recommendation is increasingly leveraging their strengths in semantic understanding and contextual reasoning. However, LM representations are inherently optimized for general semantic tasks, while recommendation models rely heavily on sparse user/item unique identity (ID) features. Existing works overlook the fundamental representational divergence between large models and recommendation systems, resulting in incompatible multimodal representations and suboptimal recommendation performance. To bridge this gap, we propose RecGOAT, a novel yet simple dual semantic alignment framework for LLM-enhanced multimodal recommendation, which offers theoretically guaranteed alignment capability. RecGOAT first employs graph attention networks to enrich collaborative semantics by modeling item-item, user-item, and user-user relationships, leveraging user/item LM representations and interaction history. Furthermore, we design a dual-granularity progressive multimodality-ID alignment framework, which achieves instance-level and distribution-level semantic alignment via cross-modal contrastive learning (CMCL) and optimal adaptive transport (OAT), respectively. Theoretically, we demonstrate that the unified representations derived from our alignment framework exhibit superior semantic consistency and comprehensiveness. Extensive experiments on three public benchmarks show that our RecGOAT achieves state-of-the-art performance, empirically validating our theoretical insights. Additionally, the deployment on a large-scale online advertising platform confirms the model's effectiveness and scalability in industrial recommendation scenarios. Code available at https://github.com/6lyc/RecGOAT-LLM4Rec.

Audio-to-image retrieval offers an interpretable alternative to audio-only classification for bioacoustic species recognition, but learning aligned audio-image representations is challenging due to the scarcity of paired audio-image data. We propose a simple and data-efficient approach that enables audio-to-image retrieval without any audio-image supervision. Our proposed method uses text as a semantic intermediary: we distill the text embedding space of a pretrained image-text model (BioCLIP-2), which encodes rich visual and taxonomic structure, into a pretrained audio-text model (BioLingual) by fine-tuning its audio encoder with a contrastive objective. This distillation transfers visually grounded semantics into the audio representation, inducing emergent alignment between audio and image embeddings without using images during training. We evaluate the resulting model on multiple bioacoustic benchmarks. The distilled audio encoder preserves audio discriminative power while substantially improving audio-text alignment on focal recordings and soundscape datasets. Most importantly, on the SSW60 benchmark, the proposed approach achieves strong audio-to-image retrieval performance exceeding baselines based on zero-shot model combinations or learned mappings between text embeddings, despite not training on paired audio-image data. These results demonstrate that indirect semantic transfer through text is sufficient to induce meaningful audio-image alignment, providing a practical solution for visually grounded species recognition in data-scarce bioacoustic settings.

The advancement of data-driven artificial intelligence (AI), particularly machine learning, heavily depends on large-scale benchmarks. Despite remarkable progress across domains ranging from pattern recognition to intelligent decision-making in recent decades, exemplified by breakthroughs in board games, card games, and electronic sports games, there remains a pressing need for more challenging benchmarks to drive further research. To this end, this paper proposes OpenGuanDan, a novel benchmark that enables both efficient simulation of GuanDan (a popular four-player, multi-round Chinese card game) and comprehensive evaluation of both learning-based and rule-based GuanDan AI agents. OpenGuanDan poses a suite of nontrivial challenges, including imperfect information, large-scale information set and action spaces, a mixed learning objective involving cooperation and competition, long-horizon decision-making, variable action spaces, and dynamic team composition. These characteristics make it a demanding testbed for existing intelligent decision-making methods. Moreover, the independent API for each player allows human-AI interactions and supports integration with large language models. Empirically, we conduct two types of evaluations: (1) pairwise competitions among all GuanDan AI agents, and (2) human-AI matchups. Experimental results demonstrate that while current learning-based agents substantially outperform rule-based counterparts, they still fall short of achieving superhuman performance, underscoring the need for continued research in multi-agent intelligent decision-making domain. The project is publicly available at https://github.com/GameAI-NJUPT/OpenGuanDan.

Research in time series anomaly detection (TSAD) has largely focused on developing increasingly sophisticated, hard-to-train, and expensive-to-infer neural architectures. We revisit this paradigm and show that a simple linear autoregressive anomaly score with the closed-form solution provided by ordinary least squares (OLS) regression consistently matches or outperforms state-of-the-art deep detectors. From a theoretical perspective, we show that linear models capture a broad class of anomaly types, estimating a finite-history Gaussian process conditional density. From a practical side, across extensive univariate and multivariate benchmarks, the proposed approach achieves superior accuracy while requiring orders of magnitude fewer computational resources. Thus, future research should consistently include strong linear baselines and, more importantly, develop new benchmarks with richer temporal structures pinpointing the advantages of deep learning models.

Electroencephalography (EEG) is a widely used technique for measuring brain activity. EEG-based signals can reveal a persons emotional state, as they directly reflect activity in different brain regions. Emotion-aware systems and EEG-based emotion recognition are a growing research area. This paper presents how machine learning (ML) models categorize a limited dataset of EEG signals into three different classes, namely Negative, Neutral, or Positive. It also presents the complete workflow, including data preprocessing and comparison of ML models. To understand which ML classification model works best for this kind of problem, we train and test the following three commonly used models: logistic regression (LR), support vector machine (SVM), and random forest (RF). The performance of each is evaluated with respect to accuracy and F1-score. The results indicate that ML models can be effectively utilized for three-way emotion classification of EEG signals. Among the three ML models trained on the available dataset, the RF model gave the best results. Its higher accuracy and F1-score suggest that it is able to capture the emotional patterns more accurately and effectively than the other two models. The RF model also outperformed the existing state-of-the-art classification models in terms of the accuracy parameter.

Neuropathological analyses benefit from spatially precise volumetric reconstructions that enhance anatomical delineation and improve morphometric accuracy. Our prior work has shown the feasibility of reconstructing 3D brain volumes from 2D dissection photographs. However these outputs sometimes exhibit coarse, overly smooth reconstructions of structures, especially under high anisotropy (i.e., reconstructions from thick slabs). Here, we introduce a computationally efficient super-resolution step that imputes slices to generate anatomically consistent isotropic volumes from anisotropic 3D reconstructions of dissection photographs. By training on domain-randomized synthetic data, we ensure that our method generalizes across dissection protocols and remains robust to large slab thicknesses. The imputed volumes yield improved automated segmentations, achieving higher Dice scores, particularly in cortical and white matter regions. Validation on surface reconstruction and atlas registration tasks demonstrates more accurate cortical surfaces and MRI registration. By enhancing the resolution and anatomical fidelity of photograph-based reconstructions, our approach strengthens the bridge between neuropathology and neuroimaging. Our method is publicly available at https://surfer.nmr.mgh.harvard.edu/fswiki/mri_3d_photo_recon

Zero-knowledge circuits enable privacy-preserving and scalable systems but are difficult to implement correctly due to the tight coupling between witness computation and circuit constraints. We present zkCraft, a practical framework that combines deterministic, R1CS-aware localization with proof-bearing search to detect semantic inconsistencies. zkCraft encodes candidate constraint edits into a single Row-Vortex polynomial and replaces repeated solver queries with a Violation IOP that certifies the existence of edits together with a succinct proof. Deterministic LLM-driven mutation templates bias exploration toward edge cases while preserving auditable algebraic verification. Evaluation on real Circom code shows that proof-bearing localization detects diverse under- and over-constrained faults with low false positives and reduces costly solver interaction. Our approach bridges formal verification and automated debugging, offering a scalable path for robust ZK circuit development.

Customer-service question answering (QA) systems increasingly rely on conversational language understanding. While Large Language Models (LLMs) achieve strong performance, their high computational cost and deployment constraints limit practical use in resource-constrained environments. Small Language Models (SLMs) provide a more efficient alternative, yet their effectiveness for multi-turn customer-service QA remains underexplored, particularly in scenarios requiring dialogue continuity and contextual understanding. This study investigates instruction-tuned SLMs for context-summarized multi-turn customer-service QA, using a history summarization strategy to preserve essential conversational state. We also introduce a conversation stage-based qualitative analysis to evaluate model behavior across different phases of customer-service interactions. Nine instruction-tuned low-parameterized SLMs are evaluated against three commercial LLMs using lexical and semantic similarity metrics alongside qualitative assessments, including human evaluation and LLM-as-a-judge methods. Results show notable variation across SLMs, with some models demonstrating near-LLM performance, while others struggle to maintain dialogue continuity and contextual alignment. These findings highlight both the potential and current limitations of low-parameterized language models for real-world customer-service QA systems.

Optimizing the structure of molecules to achieve desired properties is a central bottleneck across the chemical sciences, particularly in the pharmaceutical industry where it underlies the discovery of new drugs. Since molecular property evaluation often relies on costly and rate-limited oracles, such as experimental assays, molecular optimization must be highly sample-efficient. To address this, we introduce SEISMO, an LLM agent that performs strictly online, inference-time molecular optimization, updating after every oracle call without the need for population-based or batched learning. SEISMO conditions each proposal on the full optimization trajectory, combining natural-language task descriptions with scalar scores and, when available, structured explanatory feedback. Across the Practical Molecular Optimization benchmark of 23 tasks, SEISMO achieves a 2-3 times higher area under the optimisation curve than prior methods, often reaching near-maximal task scores within 50 oracle calls. Our additional medicinal-chemistry tasks show that providing explanatory feedback further improves efficiency, demonstrating that leveraging domain knowledge and structured information is key to sample-efficient molecular optimization.

In reverse osmosis desalination, ultrafiltration (UF) membranes degrade due to fouling, leading to performance loss and costly downtime. Most plants rely on scheduled preventive maintenance, since existing predictive maintenance models, often based on opaque machine learning methods, lack interpretability and operator trust. This study proposes an explainable prognostic framework for UF membrane remaining useful life (RUL) estimation using fuzzy similarity reasoning. A physics-informed Health Index, derived from transmembrane pressure, flux, and resistance, captures degradation dynamics, which are then fuzzified via Gaussian membership functions. Using a similarity measure, the model identifies historical degradation trajectories resembling the current state and formulates RUL predictions as Takagi-Sugeno fuzzy rules. Each rule corresponds to a historical exemplar and contributes to a transparent, similarity-weighted RUL estimate. Tested on 12,528 operational cycles from an industrial-scale UF system, the framework achieved a mean absolute error of 4.50 cycles, while generating interpretable rule bases consistent with expert understanding.

Kirkpatrick et al. [ALT 2019] and Fallat et al. [JMLR 2023] introduced non-clashing teaching and proved that it is the most efficient batch machine teaching model satisfying the collusion-avoidance benchmark established in the seminal work of Goldman and Mathias [COLT 1993]. Recently, (positive) non-clashing teaching was thoroughly studied for balls in graphs, yielding numerous algorithmic and combinatorial results. In particular, Chalopin et al. [COLT 2024] and Ganian et al. [ICLR 2025] gave an almost complete picture of the complexity landscape of the positive variant, showing that it is tractable only for restricted graph classes due to the non-trivial nature of the problem and concept class. In this work, we consider (positive) non-clashing teaching for closed neighborhoods in graphs. This concept class is not only extensively studied in various related contexts, but it also exhibits broad generality, as any finite binary concept class can be equivalently represented by a set of closed neighborhoods in a graph. In comparison to the works on balls in graphs, we provide improved algorithmic results, notably including FPT algorithms for more general classes of parameters, and we complement these results by deriving stronger lower bounds. Lastly, we obtain combinatorial upper bounds for wider classes of graphs.

Graph Domain Adaptation (GDA) typically uses adversarial learning to align graph embeddings in Euclidean space. However, this paradigm suffers from two critical challenges: Structural Degeneration, where hierarchical and semantic representations are entangled, and Optimization Instability, which arises from oscillatory dynamics of minimax adversarial training. To tackle these issues, we propose DisRFM, a geometry-aware GDA framework that unifies Riemannian embedding and flow-based transport. First, to overcome structural degeneration, we embed graphs into a Riemannian manifold. By adopting polar coordinates, we explicitly disentangle structure (radius) from semantics (angle). Then, we enforce topology preservation through radial Wasserstein alignment and semantic discrimination via angular clustering, thereby preventing feature entanglement and collapse. Second, we address the instability of adversarial alignment by using Riemannian flow matching. This method learns a smooth vector field to guide source features toward the target along geodesic paths, guaranteeing stable convergence. The geometric constraints further guide the flow to maintain the disentangled structure during transport. Theoretically, we prove the asymptotic stability of the flow matching and derive a tighter bound for the target risk. Extensive experiments demonstrate that DisRFM consistently outperforms state-of-the-art methods.

While existing multivariate time series forecasting models have advanced significantly in modeling periodicity, they largely neglect the periodic heterogeneity common in real-world data, where variates exhibit distinct and dynamically changing periods. To effectively capture this periodic heterogeneity, we propose PHAT (Period Heterogeneity-Aware Transformer). Specifically, PHAT arranges multivariate inputs into a three-dimensional "periodic bucket" tensor, where the dimensions correspond to variate group characteristics with similar periodicity, time steps aligned by phase, and offsets within the period. By restricting interactions within buckets and masking cross-bucket connections, PHAT effectively avoids interference from inconsistent periods. We also propose a positive-negative attention mechanism, which captures periodic dependencies from two perspectives: periodic alignment and periodic deviation. Additionally, the periodic alignment attention scores are decomposed into positive and negative components, with a modulation term encoding periodic priors. This modulation constrains the attention mechanism to more faithfully reflect the underlying periodic trends. A mathematical explanation is provided to support this property. We evaluate PHAT comprehensively on 14 real-world datasets against 18 baselines, and the results show that it significantly outperforms existing methods, achieving highly competitive forecasting performance. Our sources is available at GitHub.

Vision-language models have transformed multimodal representation learning, yet dominant contrastive approaches like CLIP require large batch sizes, careful negative sampling, and extensive hyperparameter tuning. We introduce NOVA, a NOn-contrastive Vision-language Alignment framework based on joint embedding prediction with distributional regularization. NOVA aligns visual representations to a frozen, domain-specific text encoder by predicting text embeddings from augmented image views, while enforcing an isotropic Gaussian structure via Sketched Isotropic Gaussian Regularization (SIGReg). This eliminates the need for negative sampling, momentum encoders, or stop-gradients, reducing the training objective to a single hyperparameter. We evaluate NOVA on zeroshot chest X-ray classification using ClinicalBERT as the text encoder and Vision Transformers trained from scratch on MIMIC-CXR. On zero-shot classification across three benchmark datasets, NOVA outperforms multiple standard baselines while exhibiting substantially more consistent training runs. Our results demonstrate that non-contrastive vision-language pretraining offers a simpler, more stable, and more effective alternative to contrastive methods.

With the proliferation of distributed data sources, Federated Learning (FL) has emerged as a key approach to enable collaborative intelligence through decentralized model training while preserving data privacy. However, conventional FL algorithms often suffer from performance disparities across clients caused by heterogeneous data distributions and unequal participation, which leads to unfair outcomes. Specifically, we focus on two core fairness challenges, i.e., representation bias, arising from misaligned client representations, and collaborative bias, stemming from inequitable contribution during aggregation, both of which degrade model performance and generalizability. To mitigate these disparities, we propose CoRe-Fed, a unified optimization framework that bridges collaborative and representation fairness via embedding-level regularization and fairness-aware aggregation. Initially, an alignment-driven mechanism promotes semantic consistency between local and global embeddings to reduce representational divergence. Subsequently, a dynamic reward-penalty-based aggregation strategy adjusts each client's weight based on participation history and embedding alignment to ensure contribution-aware aggregation. Extensive experiments across diverse models and datasets demonstrate that CoRe-Fed improves both fairness and model performance over the state-of-the-art baseline algorithms.

While Large Language Models (LLMs) have demonstrated impressive general capabilities, their direct application in the legal domain is often hindered by a lack of precise domain knowledge and complexity of performing rigorous multi-step judicial reasoning. To address this gap, we present LegalOne, a family of foundational models specifically tailored for the Chinese legal domain. LegalOne is developed through a comprehensive three-phase pipeline designed to master legal reasoning. First, during mid-training phase, we propose Plasticity-Adjusted Sampling (PAS) to address the challenge of domain adaptation. This perplexity-based scheduler strikes a balance between the acquisition of new knowledge and the retention of original capabilities, effectively establishing a robust legal foundation. Second, during supervised fine-tuning, we employ Legal Agentic CoT Distillation (LEAD) to distill explicit reasoning from raw legal texts. Unlike naive distillation, LEAD utilizes an agentic workflow to convert complex judicial processes into structured reasoning trajectories, thereby enforcing factual grounding and logical rigor. Finally, we implement a Curriculum Reinforcement Learning (RL) strategy. Through a progressive reinforcement process spanning memorization, understanding, and reasoning, LegalOne evolves from simple pattern matching to autonomous and reliable legal reasoning. Experimental results demonstrate that LegalOne achieves state-of-the-art performance across a wide range of legal tasks, surpassing general-purpose LLMs with vastly larger parameter counts through enhanced knowledge density and efficiency. We publicly release the LegalOne weights and the LegalKit evaluation framework to advance the field of Legal AI, paving the way for deploying trustworthy and interpretable foundation models in high-stakes judicial applications.

In this paper, we propose a general methodology for sampling from un-normalized densities defined on Riemannian manifolds, with a particular focus on multi-modal targets that remain challenging for existing sampling methods. Inspired by the framework of diffusion models developed for generative modeling, we introduce a sampling algorithm based on the simulation of a non-equilibrium deterministic dynamics that transports an easy-to-sample noise distribution toward the target. At the marginal level, the induced density path follows a prescribed stochastic interpolant between the noise and target distributions, specifically constructed to respect the underlying Riemannian geometry. In contrast to related generative modeling approaches that rely on machine learning, our method is entirely training-free. It instead builds on iterative posterior sampling procedures using only standard Monte Carlo techniques, thereby extending recent diffusion-based sampling methodologies beyond the Euclidean setting. We complement our approach with a rigorous theoretical analysis and demonstrate its effectiveness on a range of multi-modal sampling problems, including high-dimensional and heavy-tailed examples.

Bayesian optimization (BO) has been widely used to optimize expensive and black-box functions across various domains. Existing BO methods have not addressed tensor-output functions. To fill this gap, we propose a novel tensor-output BO method. Specifically, we first introduce a tensor-output Gaussian process (TOGP) with two classes of tensor-output kernels as a surrogate model of the tensor-output function, which can effectively capture the structural dependencies within the tensor. Based on it, we develop an upper confidence bound (UCB) acquisition function to select the queried points. Furthermore, we introduce a more complex and practical problem setting, named combinatorial bandit Bayesian optimization (CBBO), where only a subset of the outputs can be selected to contribute to the objective function. To tackle this, we propose a tensor-output CBBO method, which extends TOGP to handle partially observed outputs, and accordingly design a novel combinatorial multi-arm bandit-UCB2 (CMAB-UCB2) criterion to sequentially select both the queried points and the optimal output subset. Theoretical regret bounds for the two methods are established, ensuring their sublinear performance. Extensive synthetic and real-world experiments demonstrate their superiority.

This thesis advances semantic representation learning to render language representations or models more semantically and geometrically interpretable, and to enable localised, quasi-symbolic, compositional control through deliberate shaping of their latent space geometry. We pursue this goal within a VAE framework, exploring two complementary research directions: (i) Sentence-level learning and control: disentangling and manipulating specific semantic features in the latent space to guide sentence generation, with explanatory text serving as the testbed; and (ii) Reasoning-level learning and control: isolating and steering inference behaviours in the latent space to control NLI. In this direction, we focus on Explanatory NLI tasks, in which two premises (explanations) are provided to infer a conclusion. The overarching objective is to move toward language models whose internal semantic representations can be systematically interpreted, precisely structured, and reliably directed. We introduce a set of novel theoretical frameworks and practical methodologies, together with corresponding experiments, to demonstrate that our approaches enhance both the interpretability and controllability of latent spaces for natural language across the thesis.

Ensuring the safety of environmental exploration is a critical problem in reinforcement learning (RL). While limiting exploration to a feasible zone has become widely accepted as a way to ensure safety, key questions remain unresolved: what is the maximum feasible zone achievable through exploration, and how can it be identified? This paper, for the first time, answers these questions by revealing that the goal of safe exploration is to find the equilibrium between the feasible zone and the environment model. This conclusion is based on the understanding that these two components are interdependent: a larger feasible zone leads to a more accurate environment model, and a more accurate model, in turn, enables exploring a larger zone. We propose the first equilibrium-oriented safe exploration framework called safe equilibrium exploration (SEE), which alternates between finding the maximum feasible zone and the least uncertain model. Using a graph formulation of the uncertain model, we prove that the uncertain model obtained by SEE is monotonically refined, the feasible zones monotonically expand, and both converge to the equilibrium of safe exploration. Experiments on classic control tasks show that our algorithm successfully expands the feasible zones with zero constraint violation, and achieves the equilibrium of safe exploration within a few iterations.

Existing face parsing methods usually misclassify occlusions as facial components. This is because occlusion is a high-level concept, it does not refer to a concrete category of object. Thus, constructing a real-world face dataset covering all categories of occlusion object is almost impossible and accurate mask annotation is labor-intensive. To deal with the problems, we present S$^3$POT, a contrast-driven framework synergizing face generation with self-supervised spatial prompting, to achieve occlusion segmentation. The framework is inspired by the insights: 1) Modern face generators' ability to realistically reconstruct occluded regions, creating an image that preserve facial geometry while eliminating occlusion, and 2) Foundation segmentation models' (e.g., SAM) capacity to extract precise mask when provided with appropriate prompts. In particular, S$^3$POT consists of three modules: Reference Generation (RF), Feature enhancement (FE), and Prompt Selection (PS). First, a reference image is produced by RF using structural guidance from parsed mask. Second, FE performs contrast of tokens between raw and reference images to obtain an initial prompt, then modifies image features with the prompt by cross-attention. Third, based on the enhanced features, PS constructs a set of positive and negative prompts and screens them with a self-attention network for a mask decoder. The network is learned under the guidance of three novel and complementary objective functions without occlusion ground truth mask involved. Extensive experiments on a dedicatedly collected dataset demonstrate S$^3$POT's superior performance and the effectiveness of each module.

Most existing offline RL methods presume the availability of action labels within the dataset, but in many practical scenarios, actions may be missing due to privacy, storage, or sensor limitations. We formalise the setting of action-free offline-to-online RL, where agents must learn from datasets consisting solely of $(s,r,s')$ tuples and later leverage this knowledge during online interaction. To address this challenge, we propose learning state policies that recommend desirable next-state transitions rather than actions. Our contributions are twofold. First, we introduce a simple yet novel state discretisation transformation and propose Offline State-Only DecQN (\algo), a value-based algorithm designed to pre-train state policies from action-free data. \algo{} integrates the transformation to scale efficiently to high-dimensional problems while avoiding instability and overfitting associated with continuous state prediction. Second, we propose a novel mechanism for guided online learning that leverages these pre-trained state policies to accelerate the learning of online agents. Together, these components establish a scalable and practical framework for leveraging action-free datasets to accelerate online RL. Empirical results across diverse benchmarks demonstrate that our approach improves convergence speed and asymptotic performance, while analyses reveal that discretisation and regularisation are critical to its effectiveness.

We investigate the extent to which an LLM's hidden-state geometry can be recovered from its behavior in psycholinguistic experiments. Across eight instruction-tuned transformer models, we run two experimental paradigms -- similarity-based forced choice and free association -- over a shared 5,000-word vocabulary, collecting 17.5M+ trials to build behavior-based similarity matrices. Using representational similarity analysis, we compare behavioral geometries to layerwise hidden-state similarity and benchmark against FastText, BERT, and cross-model consensus. We find that forced-choice behavior aligns substantially more with hidden-state geometry than free association. In a held-out-words regression, behavioral similarity (especially forced choice) predicts unseen hidden-state similarities beyond lexical baselines and cross-model consensus, indicating that behavior-only measurements retain recoverable information about internal semantic geometry. Finally, we discuss implications for the ability of behavioral tasks to uncover hidden cognitive states.

Multivariate long-term time series forecasting (LTSF) supports critical applications such as traffic-flow management, solar-power scheduling, and electricity-transformer monitoring. The existing LTSF paradigms follow a three-stage pipeline of embedding, backbone refinement, and long-horizon prediction. However, the behaviors of individual backbone layers remain underexplored. We introduce layer sensitivity, a gradient-based metric inspired by GradCAM and effective receptive field theory, which quantifies both positive and negative contributions of each time point to a layer's latent features. Applying this metric to a three-layer MLP backbone reveals depth-specific specialization in modeling temporal dynamics in the input sequence. Motivated by these insights, we propose MoDEx, a lightweight Mixture of Depth-specific Experts, which replaces complex backbones with depth-specific MLP experts. MoDEx achieves state-of-the-art accuracy on seven real-world benchmarks, ranking first in 78 percent of cases, while using significantly fewer parameters and computational resources. It also integrates seamlessly into transformer variants, consistently boosting their performance and demonstrating robust generalizability as an efficient and high-performance LTSF framework.

The zero-shot evaluation of time series foundation models (TSFMs) for classification typically uses a frozen encoder followed by a task-specific classifier. However, this practice violates the training-free premise of zero-shot deployment and introduces evaluation bias due to classifier-dependent training choices. To address this issue, we propose TIC-FM, an in-context learning framework that treats the labeled training set as context and predicts labels for all test instances in a single forward pass, without parameter updates. TIC-FM pairs a time series encoder and a lightweight projection adapter with a split-masked latent memory Transformer. We further provide theoretical justification that in-context inference can subsume trained classifiers and can emulate gradient-based classifier training within a single forward pass. Experiments on 128 UCR datasets show strong accuracy, with consistent gains in the extreme low-label situation, highlighting training-free transfer

Safety alignment in Large Language Models (LLMs) often creates a systematic discrepancy between a model's aligned output and the underlying pre-aligned data distribution. We propose a framework in which the effect of safety alignment on next-token prediction is modeled as a systematic distortion of a pre-alignment distribution. We cast Weak-to-Strong Jailbreaking as a forecast aggregation problem and derive an optimal aggregation strategy characterized by a Gradient Shift in the loss-induced dual space. We show that logit-arithmetic jailbreaking methods are a special case of this framework under cross-entropy loss, and derive a broader family of aggregation rules corresponding to other proper losses. We also propose a new hybrid aggregation rule. Evaluations across red-teaming benchmarks and math utility tasks using frontier models demonstrate that our approach achieves superior Attack Success Rates and lower "Jailbreak Tax" compared with existing methods, especially on the safety-hardened gpt-oss-120b.

Text-to-Image (T2I) diffusion models enable high-quality open-ended synthesis, but their real-world deployment demands safeguards that suppress unsafe generations without degrading benign prompt-image alignment. We formalize this tension through a total variation (TV) lens: once the reference conditional distribution is fixed, any nontrivial reduction in unsafe generations necessarily incurs TV deviation from the reference, yielding a principled Safety-Prompt Alignment Trade-off (SPAT). Guided by this view, we propose an inference-only prompt projection framework that selectively intervenes on high-risk prompts via a surrogate objective with verification, mapping them into a tolerance-controlled safe set while leaving benign prompts effectively unchanged, without retraining or fine-tuning the generator. Across four datasets and three diffusion backbones, our approach achieves 16.7-60.0% relative reductions in inappropriate percentage (IP) versus strong model-level alignment baselines, while preserving benign prompt-image alignment on COCO near the unaligned reference.

This paper describes a system that has been submitted to the "PolyHope-M" at RANLP2025. In this work various transformers have been implemented and evaluated for hope speech detection for English and Germany. RoBERTa has been implemented for English, while the multilingual model XLM-RoBERTa has been implemented for both English and German languages. The proposed system using RoBERTa reported a weighted f1-score of 0.818 and an accuracy of 81.8% for English. On the other hand, XLM-RoBERTa achieved a weighted f1-score of 0.786 and an accuracy of 78.5%. These results reflects the importance of improvement of pre-trained large language models and how these models enhancing the performance of different natural language processing tasks.

Diffusion Large Language Models (dLLMs) have demonstrated promising generative capabilities and are increasingly used to produce formal languages defined by context-free grammars, such as source code and chemical expressions. However, as probabilistic models, they still struggle to generate syntactically valid outputs reliably. A natural and promising direction to address this issue is to adapt constrained decoding techniques to enforce grammatical correctness during generation. However, applying these techniques faces two primary obstacles. On the one hand, the non-autoregressive nature of dLLMs renders most existing constrained decoding approaches inapplicable. On the other hand, current approaches specifically designed for dLLMs may allow intermediate outputs that are impossible to complete into valid sentences, which significantly limits their reliability in practice. To address these challenges, we present LAVE, a constrained decoding approach specifically designed for dLLMs. Our approach leverages a key property of dLLMs, namely their ability to predict token distributions for all positions in parallel during each forward pass. Whenever a new token is proposed by model, LAVE performs lookahead using these distributions to efficiently and reliably verify the validity of the proposed token. This design ensures reliable constraints by reliably preserving the potential for intermediate outputs to be extended into valid sentences. Extensive experiments across four widely used dLLMs and three representative benchmarks demonstrate that LAVE consistently outperforms existing baselines and achieves substantial improvements in syntactic correctness, while incurring negligible runtime overhead.

Embodied AI requires agents to understand goals, plan actions, and execute tasks in simulated environments.We present a comprehensive evaluation of Large Language Models (LLMs) on the VirtualHome benchmark using the Embodied Agent Interface (EAI) framework.We compare two representative 7B-parameter models OPENPANGU-7B and QWEN2.5-7B across four fundamental tasks: Goal Interpretation, Action Sequencing, Subgoal Decomposition, and Transition Modeling.We propose Structured Self-Consistency (SSC), an enhanced decoding strategy that leverages multiple sampling with domain-specific voting mechanisms to improve output quality for structured generation tasks. Experimental results demonstrate that SSC significantly enhances performance, with OPENPANGU-7B excelling at hierarchical planning while QWEN2.5-7B show advantages in action-level tasks. Our analysis reveals complementary strengths across model types, providing insights for future embodied AI system development.

Real-time generative game engines represent a paradigm shift in interactive simulation, promising to replace traditional graphics pipelines with neural world models. However, existing approaches are fundamentally constrained by the ``Memory Wall,'' restricting practical deployments to low resolutions (e.g., $64 \times 64$). This paper bridges the gap between generative models and high-resolution neural simulations by introducing a scalable \textit{Hardware-Algorithm Co-Design} framework. We identify that high-resolution generation suffers from a critical resource mismatch: the World Model is compute-bound while the Decoder is memory-bound. To address this, we propose a heterogeneous architecture that intelligently decouples these components across a cluster of AI accelerators. Our system features three core innovations: (1) an asymmetric resource allocation strategy that optimizes throughput under sequence parallelism constraints; (2) a memory-centric operator fusion scheme that minimizes off-chip bandwidth usage; and (3) a manifold-aware latent extrapolation mechanism that exploits temporal redundancy to mask latency. We validate our approach on a cluster of programmable AI accelerators, enabling real-time generation at $720 \times 480$ resolution -- a $50\times$ increase in pixel throughput over prior baselines. Evaluated on both continuous 3D racing and discrete 2D platformer benchmarks, our system delivers fluid 26.4 FPS and 48.3 FPS respectively, with an amortized effective latency of 2.7 ms. This work demonstrates that resolving the ``Memory Wall'' via architectural co-design is not merely an optimization, but a prerequisite for enabling high-fidelity, responsive neural gameplay.

We propose a novel independent and payoff-based learning framework for stochastic games that is model-free, game-agnostic, and gradient-free. The learning dynamics follow a best-response-type actor-critic architecture, where agents update their strategies (actors) using feedback from two distinct critics: a fast critic that intuitively responds to observed payoffs under limited information, and a slow critic that deliberatively approximates the solution to the underlying dynamic programming problem. Crucially, the learning process relies on non-equilibrium adaptation through smoothed best responses to observed payoffs. We establish convergence to (approximate) equilibria in two-agent zero-sum and multi-agent identical-interest stochastic games over an infinite horizon. This provides one of the first payoff-based and fully decentralized learning algorithms with theoretical guarantees in both settings. Empirical results further validate the robustness and effectiveness of the proposed approach across both classes of games.

The class of direct preference optimization (DPO) algorithms has emerged as a promising approach for solving the alignment problem in foundation models. These algorithms work with very limited feedback in the form of pairwise preferences and fine-tune models to align with these preferences without explicitly learning a reward model. While the form of feedback used by these algorithms makes the data collection process easy and relatively more accurate, its ambiguity in terms of the quality of responses could have negative implications. For example, it is not clear if a decrease (increase) in the likelihood of preferred (dispreferred) responses during the execution of these algorithms could be interpreted as a positive or negative phenomenon. In this paper, we study how to design algorithms that can leverage additional information in the form of rating gap, which informs the learner how much the chosen response is better than the rejected one. We present new algorithms that can achieve faster statistical rates than DPO in presence of accurate rating gap information. Moreover, we theoretically prove and empirically show that the performance of our algorithms is robust to inaccuracy in rating gaps. Finally, we demonstrate the solid performance of our methods in comparison to a number of DPO-style algorithms across a wide range of LLMs and evaluation benchmarks.

Interlingual subtitling, which translates subtitles of visual media into a target language, is essential for entertainment localization but has not yet been explored in machine translation. Although Large Language Models (LLMs) have significantly advanced the general capabilities of machine translation, the distinctive characteristics of subtitle texts pose persistent challenges in interlingual subtitling, particularly regarding semantic coherence, pronoun and terminology translation, and translation expressiveness. To address these issues, we present Hermes, an LLM-based automated subtitling framework. Hermes integrates three modules: Speaker Diarization, Terminology Identification, and Expressiveness Enhancement, which effectively tackle the above challenges. Experiments demonstrate that Hermes achieves state-of-the-art diarization performance and generates expressive, contextually coherent translations, thereby advancing research in interlingual subtitling.

Temporal Graph Neural Networks (TGNNs) aim to capture the evolving structure and timing of interactions in dynamic graphs. Although many models incorporate time through encodings or architectural design, they often compute attention over entangled node and edge representations, failing to reflect their distinct temporal behaviors. Node embeddings evolve slowly as they aggregate long-term structural context, while edge features reflect transient, timestamped interactions (e.g. messages, trades, or transactions). This mismatch results in semantic attention blurring, where attention weights cannot distinguish between slowly drifting node states and rapidly changing, information-rich edge interactions. As a result, models struggle to capture fine-grained temporal dependencies and provide limited transparency into how temporal relevance is computed. This paper introduces KEAT (Kernelized Edge Attention for Temporal Graphs), a novel attention formulation that modulates edge features using a family of continuous-time kernels, including Laplacian, RBF, and learnable MLP variant. KEAT preserves the distinct roles of nodes and edges, and integrates seamlessly with both Transformer-style (e.g., DyGFormer) and message-passing (e.g., TGN) architectures. It achieves up to 18% MRR improvement over the recent DyGFormer and 7% over TGN on link prediction tasks, enabling more accurate, interpretable and temporally aware message passing in TGNNs.

A good language model starts with a good tokenizer. Tokenization is especially important for speech modeling, which must handle continuous signals that mix linguistic and non-linguistic information. A speech tokenizer should extract phonetics and prosody, suppress linguistically irrelevant information like speaker identity, and enable high-quality synthesis. We present Kanade, a single-layer disentangled speech tokenizer that realizes this ideal. Kanade separates out acoustic constants to create a single stream of tokens that captures rich phonetics and prosody. It does so without the need for auxiliary methods that existing disentangled codecs often rely on. Experiments show that Kanade achieves state-of-the-art speaker disentanglement and lexical availability, while maintaining excellent reconstruction quality.

Despite progress on general tasks, VLMs struggle with challenges demanding both detailed visual grounding and deliberate knowledge-based reasoning, a synergy not captured by existing benchmarks that evaluate these skills separately. To close this gap, we introduce Pix2Fact, a new visual question-answering benchmark designed to evaluate expert-level perception and knowledge-intensive multi-hop reasoning. Pix2Fact contains 1,000 high-resolution (4K+) images spanning 8 daily-life scenarios and situations, with questions and answers meticulously crafted by annotators holding PhDs from top global universities working in partnership with a professional data annotation firm. Each question requires detailed visual grounding, multi-hop reasoning, and the integration of external knowledge to answer. Our evaluation of 9 state-of-the-art VLMs, including proprietary models like Gemini-3-Pro and GPT-5, reveals the substantial challenge posed by Pix2Fact: the most advanced model achieves only 24.0% average accuracy, in stark contrast to human performance of 56%. This significant gap underscores the limitations of current models in replicating human-level visual comprehension. We believe Pix2Fact will serve as a critical benchmark to drive the development of next-generation multimodal agents that combine fine-grained perception with robust, knowledge-based reasoning.

Reliable Docker-based environment construction is a dominant bottleneck for scaling execution-grounded training and evaluation of software engineering agents. We introduce DockSmith, a specialized agentic Docker builder designed to address this challenge. DockSmith treats environment construction not only as a preprocessing step, but as a core agentic capability that exercises long-horizon tool use, dependency reasoning, and failure recovery, yielding supervision that transfers beyond Docker building itself. DockSmith is trained on large-scale, execution-grounded Docker-building trajectories produced by a SWE-Factory-style pipeline augmented with a loop-detection controller and a cross-task success memory. Training a 30B-A3B model on these trajectories achieves open-source state-of-the-art performance on Multi-Docker-Eval, with 39.72% Fail-to-Pass and 58.28% Commit Rate. Moreover, DockSmith improves out-of-distribution performance on SWE-bench Verified, SWE-bench Multilingual, and Terminal-Bench 2.0, demonstrating broader agentic benefits of environment construction.

Understanding structure-property relationships in complex materials requires integrating complementary measurements across multiple length scales. Here we propose an interpretable "multimodal" machine learning framework that unifies heterogeneous analytical systems for end-to-end characterization, demonstrated on carbon nanotube (CNT) films whose properties are highly sensitive to microstructural variations. Quantitative morphology descriptors are extracted from SEM images via binarization, skeletonization, and network analysis, capturing curvature, orientation, intersection density, and void geometry. These SEM-derived features are fused with Raman indicators of crystallinity/defect states, specific surface area from gas adsorption, and electrical surface resistivity. Multi-dimensional visualization using radar plots and UMAP reveals clear clustering of CNT films according to crystallinity and entanglements. Regression models trained on the multimodal feature set show that nonlinear approaches, particularly XGBoost, achieve the best predictive accuracy under leave-one-out cross-validation. Feature-importance analysis further provides physically meaningful interpretations: surface resistivity is primarily governed by junction-to-junction transport length scales, crystallinity/defect-related metrics, and network connectivity, whereas specific surface area is dominated by intersection density and void size. The proposed multimodal machine learning framework offers a general strategy for data-driven, explainable characterization of complex materials.

Time series forecasting is important in many fields that require accurate predictions for decision-making. Patching techniques, commonly used and effective in time series modeling, help capture temporal dependencies by dividing the data into patches. However, existing patch-based methods fail to dynamically select patches and typically use all patches during the prediction process. In real-world time series, there are often low-quality issues during data collection, such as missing values, distribution shifts, anomalies and white noise, which may cause some patches to contain low-quality information, negatively impacting the prediction results. To address this issue, this study proposes a robust time series forecasting framework called SEER. Firstly, we propose an Augmented Embedding Module, which improves patch-wise representations using a Mixture-of-Experts (MoE) architecture and obtains series-wise token representations through a channel-adaptive perception mechanism. Secondly, we introduce a Learnable Patch Replacement Module, which enhances forecasting robustness and model accuracy through a two-stage process: 1) a dynamic filtering mechanism eliminates negative patch-wise tokens; 2) a replaced attention module substitutes the identified low-quality patches with global series-wise token, further refining their representations through a causal attention mechanism. Comprehensive experimental results demonstrate the SOTA performance of SEER.

This paper investigates the changing nature of French drama between 1700-1900 using Latent Dirichlet Allocation and Jensen-Shannon Divergence. Results indicate that the topical distribution of French drama changed profoundly after the French Revolution, particularly between 1789 and 1850. Bourgeois themes emerged among the most prevalent topics since the late 18th century. To assess the coevolution of drama and economic growth, I plot the yearly prevalence of topics alongside French GDP between 1700-1900, and discuss these changes in light of the political and economic changes prompted by the French Revolution and the industrialization of the country.

Balancing reward and safety in constrained reinforcement learning remains challenging due to poor generalization from sharp value minima and inadequate handling of heavy-tailed risk distribution. We introduce Safe Langevin Soft Actor-Critic (SL-SAC), a principled algorithm that addresses both issues through parameter-space exploration and distributional risk control. Our approach combines three key mechanisms: (1) Adaptive Stochastic Gradient Langevin Dynamics (aSGLD) for reward critics, promoting ensemble diversity and escape from poor optima; (2) distributional cost estimation via Implicit Quantile Networks (IQN) with Conditional Value-at-Risk (CVaR) optimization for tail-risk mitigation; and (3) a reactive Lagrangian relaxation scheme that adapts constraint enforcement based on the empirical CVaR of episodic costs. We provide theoretical guarantees on CVaR estimation error and demonstrate that CVaR-based Lagrange updates yield stronger constraint violation signals than expected-cost updates. On Safety-Gymnasium benchmarks, SL-SAC achieves the lowest cost in 7 out of 10 tasks while maintaining competitive returns, with cost reductions of 19-63% in velocity tasks compared to state-of-the-art baselines.

Network topology excels at structural predictions but fails to capture functional semantics encoded in biomedical literature. We present a retrieval-augmented generation (RAG) embedding framework that integrates graph neural network representations with dynamically retrieved literature-derived knowledge through contrastive learning. Benchmarking against ten embedding methods reveals task-specific complementarity: topology-focused methods achieve near-perfect link prediction (GCN: 0.983 AUROC), while RAG-GNN is the only method achieving positive silhouette scores for functional clustering (0.001 vs. negative scores for all baselines). Information-theoretic decomposition shows network topology contributes 77.3% of predictive information, while retrieved documents provide 8.6% unique information. Applied to cancer signaling networks (379 proteins, 3,498 interactions), the framework identifies DDR1 as a therapeutic target based on retrieved evidence of synthetic lethality with KRAS mutations. These results establish that topology-only and retrieval-augmented approaches serve complementary purposes: structural prediction tasks are solved by network topology alone, while functional interpretation uniquely benefits from retrieved knowledge.

Information-seeking agents have emerged as a powerful paradigm for solving knowledge-intensive tasks. Existing information-seeking agents are typically specialized for open web, documents, or local knowledge bases, which constrains scalability and cross-domain generalization. In this work, we investigate how to consolidate heterogeneous information-seeking agents into a single foundation agentic model. We study two complementary consolidation strategies: data-level consolidation, which jointly trains a unified model on a mixture of domain-specific datasets, and parameter-level consolidation, which merges independently trained agent models at the parameter level. Our analysis compares these approaches in terms of performance retention, cross-domain generalization, and interference across information-seeking behaviors. Our results show that data-level consolidation remains a strong and stable baseline, while parameter-level consolidation offers a promising, efficient alternative but suffers from interference and robustness challenges. We further identify key design factors for effective agent consolidation at the parameter level, including fine-grained merging granularity, awareness of task heterogeneity, and principled consensus strategy.

The lack of large-scale, demographically diverse face images with precise Action Unit (AU) occurrence and intensity annotations has long been recognized as a fundamental bottleneck in developing generalizable AU recognition systems. In this paper, we propose MAUGen, a diffusion-based multi-modal framework that jointly generates a large collection of photorealistic facial expressions and anatomically consistent AU labels, including both occurrence and intensity, conditioned on a single descriptive text prompt. Our MAUGen involves two key modules: (1) a Multi-modal Representation Learning (MRL) module that captures the relationships among the paired textual description, facial identity, expression image, and AU activations within a unified latent space; and (2) a Diffusion-based Image label Generator (DIG) that decodes the joint representation into aligned facial image-label pairs across diverse identities. Under this framework, we introduce Multi-Identity Facial Action (MIFA), a large-scale multimodal synthetic dataset featuring comprehensive AU annotations and identity variations. Extensive experiments demonstrate that MAUGen outperforms existing methods in synthesizing photorealistic, demographically diverse facial images along with semantically aligned AU labels.

Irregular multivariate time series forecasting (IMTSF) is challenging due to non-uniform sampling and variable asynchronicity. These irregularities violate the equidistant assumptions of standard models, hindering local temporal modeling and rendering classical frequency-domain methods ineffective for capturing global periodic structures. To address this challenge, we propose TFMixer, a joint time-frequency modeling framework for IMTS forecasting. Specifically, TFMixer incorporates a Global Frequency Module that employs a learnable Non-Uniform Discrete Fourier Transform (NUDFT) to directly extract spectral representations from irregular timestamps. In parallel, the Local Time Module introduces a query-based patch mixing mechanism to adaptively aggregate informative temporal patches and alleviate information density imbalance. Finally, TFMixer fuses the time-domain and frequency-domain representations to generate forecasts and further leverages inverse NUDFT for explicit seasonal extrapolation. Extensive experiments on real-world datasets demonstrate the state--of-the-art performance of TFMixer.

The Traveling Salesman Problem (TSP) is one of the most representative NP-hard problems in route planning and a long-standing benchmark in combinatorial optimization. Traditional heuristic tour constructors, such as Farthest or Nearest Insertion, are computationally efficient and highly practical, but their deterministic behavior limits exploration and often leads to local optima. In contrast, neural-based heuristic tour constructors alleviate this issue through guided-sampling and typically achieve superior solution quality, but at the cost of extensive training and reliance on ground-truth supervision, hindering their practical use. To bridge this gap, we propose TSP-MDF, a novel instance modification framework that equips traditional deterministic heuristic tour constructors with guided-sampling capability. Specifically, TSP-MDF introduces a neural-based instance modifier that strategically shifts node coordinates to sample multiple modified instances, on which the base traditional heuristic tour constructor constructs tours that are mapped back to the original instance, allowing traditional tour constructors to explore higher-quality tours and escape local optima. At the same time, benefiting from our instance modification formulation, the neural-based instance modifier can be trained efficiently without any ground-truth supervision, ensuring the framework maintains practicality. Extensive experiments on large-scale TSP benchmarks and real-world benchmarks demonstrate that TSP-MDF significantly improves the performance of traditional heuristics tour constructors, achieving solution quality comparable to neural-based heuristic tour constructors, but with an extremely short training time.

Large Language Models (LLMs) inevitably memorize sensitive information during training, posing significant privacy risks. Machine unlearning has emerged as a promising solution to selectively remove such information without full retraining. However, existing methods are designed for dense models and overlook model sparsification-an essential technique for efficient LLM deployment. We find that unlearning effectiveness degrades substantially on sparse models. Through empirical analysis, we reveal that this degradation occurs because existing unlearning methods require updating all parameters, yet sparsification prunes substantial weights to zero, fundamentally limiting the model's forgetting capacity. To address this challenge, we propose Sparsity-Aware Unlearning (SAU), which decouples unlearning from sparsification objectives through gradient masking that redirects updates to surviving weights, combined with importance-aware redistribution to compensate for pruned parameters. Extensive experiments demonstrate that SAU significantly outperforms existing methods on sparse LLMs, achieving effective forgetting while preserving model utility.

Can modifying the training data distribution guide optimizers toward solutions with improved generalization when training large language models (LLMs)? In this work, we theoretically analyze an in-context linear regression model with multi-head linear self-attention, and compare the training dynamics of two gradient based optimizers, namely gradient descent (GD) and sharpness-aware minimization (SAM), the latter exhibiting superior generalization properties but is prohibitively expensive for training even medium-sized LLMs. We show, for the first time, that SAM induces a lower simplicity bias (SB)-the tendency of an optimizer to preferentially learn simpler features earlier in training-and identify this reduction as a key factor underlying its improved generalization performance. Motivated by this insight, we demonstrate that altering the training data distribution by upsampling or augmenting examples learned later in training similarly reduces SB and leads to improved generalization. Our extensive experiments show that our strategy improves the performance of multiple LLMs-including Phi2-2.7B , Llama3.2-1B, Gemma3-1B-PT, and Qwen3-0.6B-Base-achieving relative accuracy gains up to 18% when fine-tuned with AdamW and Muon on mathematical reasoning tasks.

We study how to extend chain-of-thought (CoT) beyond language to better handle multimodal reasoning. While CoT helps LLMs and VLMs articulate intermediate steps, its text-only form often fails on vision-intensive problems where key intermediate states are inherently visual. We introduce modal-mixed CoT, which interleaves textual tokens with compact visual sketches represented as latent embeddings. To bridge the modality gap without eroding the original knowledge and capability of the VLM, we use the VLM itself as an encoder and train the language backbone to reconstruct its own intermediate vision embeddings, to guarantee the semantic alignment of the visual latent space. We further attach a diffusion-based latent decoder, invoked by a special control token and conditioned on hidden states from the VLM. In this way, the diffusion head carries fine-grained perceptual details while the VLM specifies high-level intent, which cleanly disentangles roles and reduces the optimization pressure of the VLM. Training proceeds in two stages: supervised fine-tuning on traces that interleave text and latents with a joint next-token and latent-reconstruction objective, followed by reinforcement learning that teaches when to switch modalities and how to compose long reasoning chains. Extensive experiments across 11 diverse multimodal reasoning tasks, demonstrate that our method yields better performance than language-only and other CoT methods. Our code will be publicly released.

Class-Incremental Learning (CIL) aims to sequentially learn new classes while mitigating catastrophic forgetting of previously learned knowledge. Conventional CIL approaches implicitly assume that classes are morphologically static, focusing primarily on preserving previously learned representations as new classes are introduced. However, this assumption neglects intra-class evolution: a phenomenon wherein instances of the same semantic class undergo significant morphological transformations, such as a larva turning into a butterfly. Consequently, a model must both discriminate between classes and adapt to evolving appearances within a single class. To systematically address this challenge, we formalize Stage-Aware CIL (Stage-CIL), a paradigm in which each class is learned progressively through distinct morphological stages. To facilitate rigorous evaluation within this paradigm, we introduce the Stage-Bench, a 10-domain, 2-stages dataset and protocol that jointly measure inter- and intra-class forgetting. We further propose STAGE, a novel method that explicitly learns abstract and transferable evolution patterns within a fixed-size memory pool. By decoupling semantic identity from transformation dynamics, STAGE enables accurate prediction of future morphologies based on earlier representations. Extensive empirical evaluation demonstrates that STAGE consistently and substantially outperforms existing state-of-the-art approaches, highlighting its effectiveness in simultaneously addressing inter-class discrimination and intra-class morphological adaptation.

The deployment of quantized neural networks on edge devices, combined with privacy regulations like GDPR, creates an urgent need for machine unlearning in quantized models. However, existing methods face critical challenges: they induce forgetting by training models to memorize incorrect labels, conflating forgetting with misremembering, and employ scalar gradient reweighting that cannot resolve directional conflicts between gradients. We propose OEU, a novel Orthogonal Entropy Unlearning framework with two key innovations: 1) Entropy-guided unlearning maximizes prediction uncertainty on forgotten data, achieving genuine forgetting rather than confident misprediction, and 2) Gradient orthogonal projection eliminates interference by projecting forgetting gradients onto the orthogonal complement of retain gradients, providing theoretical guarantees for utility preservation under first-order approximation. Extensive experiments demonstrate that OEU outperforms existing methods in both forgetting effectiveness and retain accuracy.

We derive a universal relation for the critical temperatures of the $q$-state Potts model based on the counting of domain-wall microstates. By balancing interface energy against configurational entropy, we show that the critical temperature is determined by the ratio of the coordination-dependent energy cost to the logarithm of a total multiplicity factor. This factor decomposes into a lattice-topological constant, representing a projection from an underlying orthogonal Euclidean space, and a term representing Markovian sampling in the $q$-dimensional state space. The framework recovers exact solutions for two-dimensional square, triangular, and honeycomb lattices and achieves sub-3\% accuracy for three-dimensional simple cubic, bcc, fcc, and diamond geometries. This approach unifies the Potts universality class into a single geometric classification, revealing that the phase transition is governed by the saturation of interface propagation through the lattice manifold and providing a predictive tool that characterizes the entire $q$-state family from a single topological calibration.

We estimate the plateau moduli of highly entangled end-pinned bead-spring polymer melts with Z = 100 and Z = 200 from the time-dependent elastic response to a step strain, which we first extrapolate to infinite time and then interpolate to zero strain. We present data for systems deformed in the melt state as well as for systems deformed at the primitive path level following the recent iPPA protocol. We observe excellent agreement between the plateau moduli obtained via the two deformation protocols and good agreement with the available experimental data for commodity polymer melts using a common mapping on the Kuhn scale.

Optical resonances in nanostructures enable strong enhancement of nonlinear processes at the nanoscale, such as second-harmonic generation (SHG), with high-$Q$ modes providing intensified light--matter interactions and sharp spectral selectivity for applications in filtering, sensing, and nonlinear spectroscopy. Thanks to the recent advances in thin-film lithium niobate (TFLN) technology, these key features can be now translated to lithium niobate for realizing novel nanoscale nonlinear optical platforms. Here, we demonstrate a large-area metasurface, realized by scalable nanoimprint lithography, comprising a slanted titanium dioxide (TiO$_2$) nanograting on etchless TFLN for efficient narrowband SHG. This is enabled by the optimal coupling of quasi-bound state in the continuum (q-BIC) modes with a narrowband pulsed laser pump. The demonstrated normalized SHG efficiency is $0.15\%\,\mathrm{cm}^2/\mathrm{GW}$, which is among the largest reported for LN metasurfaces. The low pump peak intensity ($3.64~\mathrm{kW}/\mathrm{cm}^2$) employed, which enables SHG even by continuous-wave pumping, allows envisioning integrated and portable photonic applications. SHG wavelength tuning from $870$ to $920~\mathrm{nm}$ with stable output power as well as polarization control is also achieved by off-normal pump illumination. This versatile platform opens new opportunities for sensing, THz generation and detection, and ultrafast electro-optic modulation of nonlinear optical signals. Se vuoi, posso anche:

Disorder in photonic crystals and waveguides creates states inside the photonic band gap. These states are often described as Lifshitz tails despite exhibiting energy distributions inconsistent with Lifshitz statistics near the band edge. Here we show that in photonic-crystal waveguides with intentionally engineered anisotropic disorder, the band-edge tail accessible experimentally follows an Urbach law universally, with cumulative statistics $F(Δ)=\exp[-(Δ/α)^β]$, where $Δ$ is the spectral detuning from the band edge, and an exponent $β\approx 1$ independent of disorder strength and orientation. In contrast to Lifshitz behavior, the density of states is maximal at the band edge and decays into the gap. Crucially, we find that the Urbach energy $α$ is anisotropic, with a pronounced directional splitting and qualitatively different scaling for disorder parallel and perpendicular to the waveguide axis. These conclusions are supported by quantitative agreement between optical measurements of GaAs photonic-crystal waveguides and full-vector simulations. The anisotropic Urbach energy emerges as a sensitive probe of disorder-mode coupling and a practical metric to characterize structural disorder in photonic devices.

Nanofluidic systems exploit nanometre-scale confinement in channels and pores to regulate ionic transport, enabling functionalities such as osmotic energy harvesting and neuromorphic ionic memory. Studying such confined transport requires both precise electrical instrumentation and careful data analysis, yet, in practice, measurements are still often taken with vendor software, exported as files, and processed later in separate environments. In this work, we bring these steps together in a unified Python-based framework built around three interoperable graphical user interfaces (GUIs) for nanochannel, nanopore and memristor experiments. The framework is organised into two functional parts, measurement and analysis. On the measurement side, two GUIs drive Keithley Source Meters to run continuous voltage sweeps and user-defined memristive pulse sequences, while providing live plots, configuration management and controlled shutdown routines. On the analysis side, a dedicated nanochannel and nanopore GUI reads raw I-V datasets, applies unit-consistent processing, extracts conductance and ion mobility, evaluates selectivity and osmotic power, and is complemented by a web-based calculator that performs the same mobility analysis without a local Python installation. All three GUIs are implemented in Python/Tkinter with modular plotting and logging layers so that flexible control sequences and physics-based post-processing share a common data format, improving reproducibility, timing stability and day-to-day efficiency in nanofluidic and electronic device studies.

Phase transitions occur when a macroscopic number of local degrees of freedom coherently change their behavior. In ground states of quantum many-body systems, phase transitions due to quantum fluctuations are observed as non-analytic behaviors of order parameters, such as magnetization, as functions of a conjugate parameter, such as the magnetic field. However, as soon as thermal fluctuations are present, these effects are believed to disappear for local observables. We show that this is not necessarily the case: order parameters may still show non-analytic behaviors within their dynamics. With the example of the Ising model and using methods based on hydrodynamic fluctuations, we evaluate the exact order-parameter correlation time, in space-time directions of all velocities, in equilibrium states at nonzero temperature. We reveal non-analytic behaviors of spin correlation times as functions of the magnetic field, velocity, and temperature. As a function of the magnetic field, they occur at values that continuously approach that of the zero-temperature equilibrium transition point as the velocity is decreased and reach it within the light cone, where we obtain a new, temperature-independent logarithmic divergence characterizing the collective dynamics. Thus, collective effects induced by quantum fluctuations persist within the dynamics of local observables.

This dissertation explores various nonlinear responses that arise from the rich interplay between quantum geometry, disorder, magnetism and topology in quantum materials. In addition to presenting generalizations of quantum kinetic theory, Kubo formulas and semiclassical Boltzmann transport theory to the nonlinear response regime, we discuss several predictions of novel transport effects and physical insights that emerge from these developments.

Dirac exceptional points (EPs) represent a novel class of non-Hermitian singularities that, unlike conventional EPs, reside entirely within the parity-time unbroken phase and exhibit linear energy dispersion. Here, we theoretically investigate the quantum geometry of Dirac EPs realized in nitrogen-vacancy centers in diamond, utilizing fidelity susceptibility as a probe. We demonstrate that despite the absence of a symmetry-breaking phase transition, the Dirac EP induces a pronounced geometric singularity, confirming the validity of fidelity in characterizing non-Hermitian EPs. Specifically, the real part of the fidelity susceptibility diverges to negative infinity, which serves as a signature of non-Hermitian criticality. Crucially, however, we reveal that this divergence exhibits a distinct anisotropy, diverging along the non-reciprocal coupling direction while remaining finite along the detuning axis. This behavior stands in stark contrast to the omnidirectional divergence observed in conventional EPs. Our findings provide a comprehensive picture of the fidelity probe near the Dirac EP, highlighting the critical role of parameter directionality in exploiting Dirac EPs for quantum control and sensing applications.

We investigate the transport properties of a triangular triple quantum dot (TTQD) system connected with two reservoirs under linear response regime. By employing the hierarchical equations of motion(HEOM), we compute the thermopower and thermoelectric figure of merit. The impact of interaction scheme among three quantum dots on the thermopower is thoroughly analyzed, while the thermoelectric current and spectral function throughout this process are also elaborated. Our results reveal that, under low-temperature conditions, the alleviation of the geometric blockade in the TTQD system leads to a significantly faster enhancement of the heat current compared to the electric current. This phenomenon consequently elevates the thermopower, resulting in a remarkably high thermoelectric figure of merit.

Quantum geometry plays a fundamental role across many branches of modern physics, yet its full characterization in nonequilibrium systems remains a challenge. Here, we propose a framework for quantum geometry in Floquet topological insulators by introducing a time-resolved quantum metric tensor, defined via the trace distance between micromotion operators in momentum-time space. For class A in two spatial dimensions, we find a general inequality linking the Floquet quantum metric tensor and the Floquet topology: the associated quantum volume is bounded below by the Floquet topological invariant. This relation is found to also hold in class AIII in one dimension, where the Floquet geometric tensor may be notably reduced due to time-reflection symmetry. This work will be useful in digesting the general aspects of quantum geometry in periodically driven systems in connection with their topological characterization.

We study a single lattice polymer chain under a fixed end-to-end distance, incorporating both Domb--Joyce (DJ) soft-core self-repulsion between polymer segments and a local bending-energy cost. By decomposing the stiffness into energetic and entropic contributions, we survey the parameter space defined by the self-repulsion strength and bending-energy cost. We find that the energetic contribution to stiffness is negative across the entire explored range, whereas the entropic contribution remains positive. These results unify two previously independent mechanisms of negative energetic elasticity -- solvent-induced self-repulsion and bending stiffness -- and demonstrate that either mechanism alone, as well as their combination, produces the same sign. Beyond this sign-level unification, we analyze the internal-energy scaling and show that, in the absence of the bending-energy term, the DJ (self-repulsion) limit exhibits a robust $(n-r)^{7/4}/n$ scaling collapse. In contrast, the introduction of finite bending stiffness progressively disrupts this scaling, providing an internal-energy-based diagnostic to distinguish between contributions from self-repulsion and bending stiffness.

We propose a general class of open fermionic models where quadratic Liouvillians governing the dissipative dynamics feature exactly solvable higher-order exceptional points (EPs). Invoking the formalism of third quantization, we show that, among the multiple EPs of Liouvillian, an EP with its order approaching the system size arises as the quasisteady state of the system, leading to a gapless Liouvillian spectrum. By introducing perturbations, in the form of many-body quantum-jump processes, these higher-order EPs break down, leading to finite Liouvillian gaps with fractional power-law scalings. While the power-law scaling is a signature of the higher-order EP, its explicit form is sensitively dependent on the many-body perturbation. Finally, we discuss the steady-state approaching dynamic which can serve as detectable signals for the higher-order Liouvillian EPs.

We study the presence of an external magnetic field, in combination with torsional strain, over the electron-phonon interactions in a type I Weyl semimetal. This particular superposition of field and strain, modeled in the continuum approximation by an effective gauge field, leads to an asymmetric pseudo-magnetic field at each Weyl node of opposite chirality. Therefore, we also studied the role of nodal asymmetry in the properties of the system by means of the Kadanoff-Wilson renormalization group and the corresponding flow equations. By solving those, we discuss the evolution of the coupling parameters of the theory, and analyze possible fixed points and lattice (Peierls) instabilities emerging from interactions between phonons with the chiral Landau level in the very strong pseudo-magnetic field regime.

In this paper, a carbon nanotube (CNT) based single-ion detector is proposed and its performance is evaluated with atomistic quantum transport models. The sensor can detect any ion type without molecule-specific functionalization and allows for continuous real-time ion monitoring. A single ion temporarily changes the operating principle of the sensor's CNT field-effect transistor into a resonant tunneling diode. The concrete device example of this paper showed a source-drain current increase of 5 orders of magnitude induced by a single ion.

Scale-invariant avalanches -- with events of all sizes following power-law distributions -- are considered critical. Above the upper critical dimension of four, the mean-field solution with a robust $3/2$ size exponent describes the dynamics. In two and three dimensions, spatial constraints yield smaller yet robust exponent values governed by universality classes. However, both earthquake data and experiments often show exponent values larger than $3/2$, challenging those theoretical arguments based on critical behavior. Through extensive simulations in the classical OFC earthquake model, here we show a clear transition from the theoretical expected behavior of a robust exponent value, to a regime of quasi-critical dynamics with larger than $3/2$ exponents that depend on dissipation. While the first critical regime exhibits an inherently memoryless behavior, both the transition and the second regime are driven by memory effects provoked by the growth of avalanches over the traces left by previous events, due to dissipative mechanisms. The system hovers at a distance $d_{cp}$ from the critical point, and accounting for a power-law distribution of $d_{cp}$, validated by susceptibility measurements, captures the transition. This framework provides a unified description of both critical and quasi-critical behavior, and thus of the full spectrum of scale-free dynamics observed in nature.

In order to solidify the usefulness of metadynamics in studying nucleation of crystals from supercooled liquids, we provide a specific procedure to calculate nucleation free energy barriers. After a pedagogical review of the important elements of classical nucleation theory and how metadynamics is used to find nucleation free energy barriers, we explain the benefits of local collective variables over more common global collective variables. We show how a metadynamics free energy barrier must be carefully postprocessed so that classical nucleation theory can be applied to calculate nucleation rates. We apply our procedure to a Yukawa plasma and show that a particular physically-motivated fit to metadynamics data reproduces low-temperature reference data, justifying the usefulness of metadynamics to predict nucleation rates and the nucleation critical temperature.

In this commentary, we clarify that the Landauer formula is not limited to the phonon gas model. It is fundamentally more general and applies to both particle- and wave-based descriptions of phonons, provided the transmission function is well defined. In the harmonic regime, the phonon transmission function and the resulting Landauer expression for heat current are exact. They can be rigorously derived using the atomistic Green's function method, which treats phonons as waves and does not require phonon dispersion in the interface region. In short, the Landauer framework remains valid for ideal, disordered, and defective interfaces, as long as an appropriate transmission function is used.

Nanoconfined water plays a crucial role in nanofluidics, biology, and cutting-edge technologies. The process of melting water monolayers and quasi-two-dimensional confined water involves, as an intermediate stage, the hexatic phase--a state that lies between solid and liquid and is characterized by quasi-long-range orientational order and short-range translational order. However, the influence of hydrogen bond (HB) cooperativity in this process has not been thoroughly investigated. This gap hampers our understanding of the phase behavior of confined water and limits the accuracy of our models. To address this, we extend the water model developed by Franzese and Stanley, which explicitly includes many-body interactions (MBIs) of HBs. We distinguish the contributions of three-body and five-body HB-MBIs. Our Monte Carlo calculations in the isobaric-isothermal ensemble produces a detailed pressure-temperature phase diagram, revealing polymorphism and polyhexaticity: low-density square ice and high-density triangular ice are separated from the liquid phase by distinct hexatic phases. Three-body interactions notably promote crystallization and can destabilize the low-density hexatic phase, while cooperative five-body interactions help restore it, thus modifying the thermodynamic landscape. These findings demonstrate that HB-MBIs are key in determining the phase behavior of confined water, influencing phenomena such as the non-monotonic specific heat, maximum density lines, and the accessibility of the liquid-liquid critical point. Beyond advancing theoretical understanding, these results have wide-ranging implications for nanofluidics, interfacial science, and applications in biology, food technology, and pharmaceutics, where controlling water under confinement is essential.

Path-integral molecular dynamics simulations and electronic structure-based energy decomposition analysis (EDA) are employed to connect hydrogen bond (H-bond) strength, its asymmetry, and the total delocalization energy at the water/air interface to experimentally measurable observables, such as the reorientation dynamics and the sum-frequency generation (SFG) spectrum. Using SFG spectra for distinct layers at the water/air interface, we validate the accuracy of our simulations and report a red-shift from the interface to bulk and a strongly bonded water peak at around 3250 cm$^{-1}$ in the layer closest to bulk. The reorientation dynamics of water molecules slow down from the interface to bulk, which correlates with the SFG results. From our EDA based on absolutely localized molecular orbitals, we observe a strong decline in total delocalization energy from bulk to the interface, as well as a decline in the strength of the strongest donor and acceptor interactions. The asymmetry between the two strongest interactions similarly rises towards the interface, while the importance of interactions from the outer solvation shells is greatly diminished and is lower than previously reported. Finally, we find that the strength of the strongest H-bond donor/acceptor is best correlated with the local minimum of the autocorrelation function resembling the L2 band librational motions. Following that, we propose a simple yet quantitative relationship between H-bond strength and the short-time reorientation dynamics at the water/air interface that could potentially be extended to predict H-bond strength in other hydrophobic systems from experimentally obtainable observables.

We propose a data-driven heuristic for NP-hard Ising and Max-Cut optimization that learns the update rule of an iterative dynamical system. The method learns a shared, node-wise update rule that maps local interaction fields to spin updates, parameterized by a compact multilayer perceptron with a small number of parameters. Training is performed using a zeroth-order optimizer, since backpropagation through long, recurrent Ising-machine dynamics leads to unstable and poorly informative gradients. We call this approach a neural network parameterized Ising machine (NPIM). Despite its low parameter count, the learned dynamics recover effective algorithmic structure, including momentum-like behavior and time-varying schedules, enabling efficient search in highly non-convex energy landscapes. Across standard Ising and neural combinatorial optimization benchmarks, NPIM achieves competitive solution quality and time-to-solution relative to recent learning-based methods and strong classical Ising-machine heuristics.

We establish a direct connection between the Feynman-Vernon path integral formalism for open quantum systems and the Wiener path integral used in classical stochastic dynamics. By considering a generalized influence functional in the strong decoherence limit, we demonstrate that integrating over the quantum coherence length leads to a derivation of stochastic Langevin dynamics. Specifically, we show that the quantum Feynman measure transforms into the stochastic Wiener measure. Applying this framework to the Wigner function representation, we show that the system follows a stochastic path interpretable via classical probability theory. Finally, we address the inverse problem: constructing an equivalent quantum influence functional from a given classical Langevin equation.

Fermionic non-Gaussianity quantifies a quantum state's deviation from a classically tractable free-fermionic description, constituting a necessary resource for computational quantum advantage. Here we use fermionic antiflatness (FAF) to measure this deviation across ergodic and many-body localized (MBL) regimes. We focus on the paradigmatic disordered spin-$1\!/2$ XXZ chain and its impurity variant with local interactions. Across highly excited eigenstates, FAF evolves from typical-state behavior at weak disorder to strongly suppressed values deep in the MBL regime, with volume-law scaling in the XXZ chain and an area-law bound in the impurity setting. Rare long range catlike eigenstates exhibit a pronounced enhancement of FAF, making it a sensitive diagnostic of mechanisms proposed to destabilize MBL. Starting from product states, we find that in the MBL regime FAF grows slowly in time, approaching saturation via a power-law relaxation. Overall, our results show that MBL suppresses fermionic non-Gaussianity, and the associated complexity beyond free fermions, while ergodicity restores it, motivating explorations of fermionic non-Gaussianity in other ergodicity-breaking phenomena.

Open quantum systems can be described by unraveling Lindblad master equations into ensembles of quantum trajectories. Here we investigate how the complexity of such trajectories is affected by conservation laws and other dynamical constraints of the underlying Lindblad evolution. We characterize this complexity using a data-driven approach based on the intrinsic dimension, defined as the minimal number of variables required to encode the information contained in a data set. Applying this framework to several systems, including dissipative variants of the quantum top and of the XXZ chain, we find that the intrinsic dimension is sensitive to the structure of their dynamics. The Lindblad evolution in these systems is typically chaotic, but additional constraints arise at specific parameter values, where the dynamics becomes integrable, exhibits Hilbert-space fragmentation, or develops a closed BBGKY hierarchy, leading to pronounced minima in the intrinsic dimension. Our approach results in an unsupervised probe of the complexity of dissipative quantum systems that is sensitive to chaos and ergodicity breaking phenomena beyond the initial transient regime.

We study tight-binding models in the crossover from hyperbolic to Euclidean lattices, realized through the successive insertion of Euclidean defects into hyperbolic lattices. We analyze how the holographic two-point boundary correlation function and bulk density of states evolve as defects are gradually introduced. We find that bulk properties are strongly affected by the presence of Euclidean defects, whereas boundary observables remain remarkably robust even at high defect fractions. This robustness indicates that essential features of boundary physics on hyperbolic lattices, which capture aspects of AdS/CFT-like dualities in discrete systems, can be reproduced both experimentally and numerically without requiring perfectly hyperbolic lattices, thereby reducing the system size needed for implementation.

We analyze mechanisms for universal out-of-equilibrium dynamics near criticality by exploring the effect of randomized quantum resetting (QR) under a finite-time quench across a quantum phase transition. Using the transverse-field Ising chain as a generic model and exploiting its exact solution, QR is found to cause a crossover of the scaling of the topological defect density with the time scale $τ$ of the quench, from Kibble-Zurek to anti-Kibble-Zurek scaling as $τ$ increases. This reflects a competition between non-adiabatic quench-driven excitations and QR, giving rise to local minima of the defect densities at optimal annealing times. These times and the corresponding local minima are shown to scale as universal power laws with the rate of QR. Additional results for the scaling of the mean excess energy suggest that a system driven across a quantum critical point exhibits the same scaling behavior under a linear quench with QR as with uncorrelated noise.

Standard treatments of quantum work using projective energy measurements erase initial coherence and alter the dynamics, thereby failing to capture the thermodynamic effects of coherent superpositions of energy eigenstates in an ensemble of initial states. In this article, we use an operational work definition that is non-intrusive, applying it to the case of a driven dissipative qubit, where the qubit's initial preparation comprises coherent superposition states, while the driving is coherence-less. We derive an evolution equation for the moment generating function for this work, faithfully capturing the thermodynamic signature of coherent superpositions in the initial ensemble. We demonstrate that different initial ensembles that correspond to the same density matrix upon ensemble average, while having the same average work, display different work fluctuations. For monotonic driving, we show that fluctuations are maximum for coherence-less initial ensembles. As an application, we consider quantum bit-erasure in finite time and demonstrate significantly different work statistics for erasing a classical bit of information versus a Haar random initial ensemble. Our results indicate that coherence in the initial ensemble can be utilized as a resource for thermodynamic precision without incurring additional dissipative work costs. We also obtain a generalized fluctuation theorem that establishes a new quantum lower bound on the mean dissipated work. This bound, counterintuitively, is also applicable to a "classical" initial ensemble with the same initial density matrix and is connected to quantum absolute irreversibility.

Odd diffusion breaks time-reversal symmetry in overdamped systems through transverse probability currents while preserving equilibrium steady states. In this work, we develop a dynamical density functional theory (DDFT) for dense interacting odd-diffusive fluids and apply it to ultrasoft particles in two dimensions. In bulk, odd diffusion qualitatively reshapes collective relaxation by generating transient circulating current patterns which do not exist in normal fluids. Under harmonic ring confinement, the circulation of probability current induces an angular redistribution of density along the ring during relaxation. This unique footprint of odd diffusion opens up a shorter pathway to equilibrium. Repulsive interactions significantly enhance these effects. Excellent agreement with Brownian dynamics simulations confirms that our odd-DDFT framework quantitatively captures all essential nonequilibrium aspects of the nontrivial odd transport and collective redistribution for dense fluids in both bulk and confined geometries.

As energy dissipation and gain are ubiquitous in the real world, such phenomena demand the generalization of Hermitian methods such as the analysis of topological properties for non-Hermitian systems. However, as non-Hermitian systems typically contain more degrees of freedom, this poses a challenge for analytical approaches to understand their topology and invariants. In this work, we analytically calculate the 2D Zak phase for a 2D non-Hermitian SSH-type Hamiltonian that supports a rich structure and edge currents. Closed-form expressions for eigenstates and divisions of the phase diagram are obtained, including for regions in the phase diagram where different types of exceptional points exist. We use Morse theory to determine the topology of exceptional points in momentum space. Although the band structure breaks down at exceptional points, we show that a specific phase-based topological invariant remains well-defined. Furthermore, our work yields an analytic derivation for counting edge states in the Hermitian limit. These results provide new conceptual and analytical tools for the study of complex topological systems.

Recent advances in coherent spin shuttling have made sparse semiconductor spin qubit arrays an appealing solid-state platform to realize quantum processors. The dynamic and long-range connectivity enabled by shuttling is also essential for many quantum error-correction (QEC) schemes. Here, we demonstrate a silicon spin-qubit device that comprises a shuttling bus for coherently transporting qubits that can interact at four isolated locations we call bus stops. We dynamically populate the array and tune all single- and two-qubit operations using shuttling and quantum non-demolition (QND) spin measurements, without access to charge sensing in most of the device. We achieve universal control of the effective five-qubit processor and select the connectivity required to form a surface-code stabilizer plaquette that supports X- and Z-type parity checks up to weight-four. We use the parity checks to generate multi-qubit entanglement between all qubit combinations in the array and report the genuine entanglement of a five-qubit Greenberger-Horne-Zeilinger (GHZ) state, constituting the largest such state ever constructed with gate-defined semiconductor spins. This work opens immediate opportunities to pursue QEC experiments with spin qubits, and the protocols developed here lay the groundwork for the modular calibration and operation of sparse spin qubit arrays.

We study the connected ensemble, a statistical-mechanics framework that characterizes the formation of low-loss paths in rugged landscapes. First introduced in a previous paper, this ensemble allows one to identify when a network can be trained on a simple task and which minima should be targeted during training. We apply this new framework to the symmetric binary perceptron model (SBP), and study how its typical {connected} minima behave. We show that {connected} minima exist only above a critical threshold $κ_{\rm connected}$, or equivalently below a critical constraint density $α_{\rm connected}$. This defines a parameter range in which training the network is easy, as local algorithms can efficiently access this connected manifold. We also highlight that these minima become increasingly robust and closer to one another as the task on which the network is trained becomes more difficult.

We investigate the dynamics of Entanglement Hamiltonians (EHs) in dissipative free-fermionic systems using a recent operator-based formulation of the quasiparticle picture. Focusing on gain and loss dissipation, we study the post-quench evolution and derive explicit expressions for the EH at the ballistic scale. In the long-time and weak-dissipation regime, the EH is shown to take the form of a time-dependent Generalized Gibbs Ensemble (t-GGE), with a structure that is universal across different initial states of the quench protocol. Within this framework, the emergence of the t-GGE is fully accounted for by the quasiparticle picture, and we argue that this description remains valid whenever the Lindbladian admits an appropriate coarse-grained representation.

We present an open-source Python package, SpinWaveToolkit (SWT), for (semi-)analytical modeling of spin-wave dynamics in thin ferromagnetic films and exchange-coupled magnetic bilayers. SWT combines analytical models based on the Kalinikos-Slavin theory with a semi-analytical dynamic-matrix approach, enabling the calculation of dispersion relations, group velocities, decay lengths, mode profiles, and static equilibrium magnetization states. In addition, SWT implements a quantitative model of micro-focused Brillouin light scattering (BLS) that incorporates vectorial optical focusing, spin-wave Bloch functions, magneto-optical coupling, and Green-function propagation to simulate experimentally measured BLS spectra. The package is validated against finite-element dynamic-matrix simulations performed with TetraX for Damon-Eshbach, backward-volume, forward-volume, and oblique-field geometries, showing excellent agreement while reducing computation times by nearly two orders of magnitude in comparison to the numerical simulations. Thanks to the easiness of the use and fast calculation times, SWT can be used not only for exploratory mapping of the parameter space, but also for the fitting of the measured dispersion relations and related parameters. Thus, it provides a versatile and efficient framework for experiment design, interpretation, and parameter optimization for magnonics research.

Mechanical strain can be used to control physical properties in materials. The experimental investigation of strain-induced effects at the nanoscale is of importance not only for its fundamental aspect, but also for the development of device applications. Transmission X-ray microscopy is a particularly well-suited technique for the nanoscale imaging of magnetic materials, but its compatibility with in-situ mechanical straining of samples is limited. In this work, we present a setup for applying tailored in-situ mechanical strains to freestanding thin films by means of a micro electromechanical system (MEMS) actuator. We then present a proof-of-concept experiment where a freestanding 80 nm thick (001) BiFeO$_3$ multiferroic thin film is strained with the MEMS device, allowing us to control the coupled ferroelectric/spin cycloidal configuration.

Magnon-photon hybrid systems consisting of a three-dimensional electromagnetic resonator and a bulk magnetic insulator constitute the standard experimental platform in cavity magnonics. Here, we demonstrate a modular loop-gap resonator design optimized to couple with thin films of magnetic insulators. We achieve the strong-coupling regime using this loop-gap resonator coupled to a 75~nm-thick epitaxial film of yttrium iron garnet at room temperature. We further show how to perform field-differential spectroscopy of the hybrid magnon-photon system, which eliminates the unwanted signal from other loop-gap modes uncoupled to the magnetic film. In addition to the uniform ferromagnetic resonance mode, the loop-gap resonator enables an hybridization with the standing spin-wave modes forming across the thickness of the film. Our approach unlocks the use of epitaxial films and multilayers of magnetic insulators to tune the magnon band structure in cavity magnonics experiments.

The quantum mechanics of the Josephson effect is the core ingredient for quantum technologies with superconducting circuits. A new avenue was recently opened in this field by predicting that the Josephson quantum mechanics in the odd parity sector, when a quasiparticle in trapped in an Andreev bound state, is fundamentally different from the conventional one in the even sector. The focus was then on a Josephson junction surrounded by an electromagnetic environment formed of a collection of bosonic modes, including the case of an ohmic environment. Here we consider the distinct case of a superconducting qubit made of a single Josephson junction whose environment reduces to a capacitance. We find a novel structure for the low-lying discrete states in the odd sector, which is altogether different from the one that appears in the even sector. Our study of the bound-state spectrum ranges from the Coulomb-dominated (Cooper pair box) to the Josephson-dominated (transmon) regime. Our prediction could be tested in forthcoming experiments with superconductor/semiconductor/superconductor junctions, which have been studied intensively in recent years, both using nanowires as well as two-dimensional electron gases.

The counterintuitive emergence of order from noise is a central phenomenon in science, ranging from pattern formation and synchronization to order-by-disorder in frustrated systems. While large-scale spatial self-organization induced by local spatial noise is well studied, whether temporal noise can also drive such organization remains an open question. Here, by studying interacting particle systems, we show that temporally correlated noise can lead to a self-organized state with suppressed long-range density fluctuations, or hyperuniformity. Further, we develop a fluctuating hydrodynamic theory that quantitatively explains the origin of this phenomenon. Finally, by casting the dynamics as a stochastic optimization problem, we show that temporal correlations lead to better solutions, akin to perturbed gradient descent in neural networks -- where noise is injected during training to escape poor minima. This reveals a striking correspondence between perturbed gradient descent dynamics on the energy landscapes of particle systems and the loss landscapes of neural networks. Our study establishes temporal correlations as a novel mechanism for noise-driven self-organization, with broad implications for self-assembling materials, biological systems, and optimization algorithms that leverage temporal noise for applications like differentially private learning.

Spider silk is a remarkable biomaterial with exceptional stiffness, strength, and toughness stemming from a unique microstructure. While recent studies show that silk fibers exhibit plasticity, hysteresis, and recovery under cyclic loading, the underlying microstructural mechanisms are not yet fully understood. In this work, we propose a mechanism explaining the loading-unloading-relaxation response through microstructural evolution: initial loading distorts intermolecular bonds, resulting in a linear elastic regime. Upon reaching the yield stress, these bonds dissociate and the external load is transferred to the polypeptide chains, which deform entropically to allow large deformations. Unloading is driven by entropic shortening until a traction free state with residual stretch is achieved. Subsequently, the fiber recovers as chains reorganize and bonds reform, locking the microstructure into a new stable equilibrium that increases stiffness in subsequent cycles. Following these mechanisms, we develop a microscopically motivated, energy-based model that captures the macroscopic response of silk fibers under cyclic loading. The response is decoupled into two parallel networks: (1) an elasto-plastic network of inter- and intramolecular bonds governing the initial stiffness and yield stress, and (2) an elastic network of entropic chains that enable large deformations. The model is validated against experimental data from Argiope bruennichi dragline silk. The findings from this work are three-fold: (1) explaining the mechanisms that govern hysteresis and recovery and linking them to microstructural evolution; (2) quantifying the recovery process of the fiber, which restores and enhances mechanical properties; and (3) establishing a predictive foundation for engineering synthetic fibers with customized properties.

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

The Physarum network expands or retracts in response to environmental stimuli, demonstrating an intelligent adaptive capability to locate optimal paths for nutrient transport. The underlying physical mechanism governing this intelligence behavior remains an unresolved problem in biological physics.unlike the unidirectional flow typical of urban traffic networks, cytoplasmic flow within the Physarum network exhibits periodic oscillations modulated by biological repellents and attractants. In this study, we investigate how local flows within the networks branch channels interact to collectively govern the global oscillatory dynamics.We find that the measured flow fluxes at intersection nodes obey Kirchhoff's current law. Phase differences exist among the flows in different branches.At the microscopic scale, flow distribution exhibits only brief periods of traffic congestion, which are resolved by the oscillatory flows. By mapping the flow flux vectors onto the magnetic moment vector of spin ice model, we demonstrate that the flow vectors strictly obey the ice-rule of vertex models in statistical physics.Notably, the three branches converging at a Y-shaped node never become blocked simultaneously, thereby preventing traffic congestion and ensuring efficient transmission of nutrients and signals.This intelligent flow control phenomenon offers novel insights for addressing traffic congestion and advances our understanding of frustrated quantum magnetism.

People learn about opportunities and actions by observing the experiences of their friends. We model how homophily -- the tendency to associate with similar others -- affects both the endogenous quality and diversity of the information accessible to decision makers. Homophily provides higher-quality information, since observing the payoffs of another person is more informative the more similar that person is to the decision maker. However, homophily can lead people to take actions that generate less information. We show how network connectivity influences the tradeoff between the endogenous quantity and quality of information. Although homophily hampers learning in sparse networks, it enhances learning in sufficiently dense networks.

While opinion dynamics models have been extensively studied as stylized models, there has been growing attention to the possibility of combining these models with empirical data. This attention seems to be driven by the many social issues that strongly depend on people's opinions (such as climate change and vaccination) and the need for empirically valid models to design related policy interventions. While different models have been combined in various ways with empirical data, standardised comparison of models against empirical data is still lacking. In this article, we test the validity of multiple opinion dynamics models--including both stylized and more realistic models. Our approach follows a "data science-like" validation procedure, where we first calibrate the model's free parameters using an initial range of years (e.g. 2010-2015), and then use data from one wave (e.g. 2016) to predict data in the following wave (e.g. 2017). We initially tested such a procedure using simulated data and then tested different models on various topics from the European Social Survey. Both toy models and empirical models perform well on the simulated data, but fail to predict future years in the empirical data. Furthermore, during the calibration phase on the empirical data, most models learned to "freeze"--meaning that their predictions for the following year are just a copy of the data from the previous year. This work advances the literature by offering a benchmark for comparing different opinion dynamics models. Furthermore, our tests show that real-world dynamics appear to be completely incompatible with the dynamics of the tested models. This calls for more effort in exploring what are the features that would improve validity and applications for opinion dynamics models.

Natural selection for terrestrial locomotion has yielded unifying patterns in the body shape of legged animals, often manifesting as scaling laws. One such pattern appears in the frontal aspect ratio. Smaller animals like insects typically adopt a landscape frontal aspect ratio, with a wider side-to-side base of support than center of mass height. Larger animals like elephants, however, are taller than wide with a portrait aspect ratio. Known explanations for postural scaling are restricted to animal groups with similar anatomical and behavioural motifs, but the trend in frontal aspect ratio transcends such commonalities. Here we show that vertebrates and invertebrates with diverse body plans, ranging in mass from 28 mg to 22000 kg, exhibit size-dependent scaling of the frontal aspect ratio driven by the need for lateral stability on uneven natural terrain. Because natural terrain exhibit scale-dependent unevenness, and the frontal aspect ratio is important for lateral stability during locomotion, smaller animals need a wider aspect ratio for stability. This prediction is based on the fractal property of natural terrain unevenness, requires no anatomical or behavioural parameters, and agrees with the measured scaling despite vast anatomical and behavioural differences. Furthermore, a statistical phylogenetic comparative analysis found that shared ancestry and random trait evolution cannot explain the measured scaling. Thus, our findings reveal that terrain roughness, acting through natural selection for stability, likely drove the macroevolution of frontal shape in terrestrial animals.

Identifying minimal and informative feature sets is a central challenge in data analysis, particularly when few data points are available. Here we present a theoretical analysis of an unsupervised feature selection pipeline based on the Differentiable Information Imbalance (DII). We consider the specific case of structural and physico-chemical features describing a set of proteins. We show that if one considers the features as coordinates of a (hypothetical) statistical physics model, this model undergoes a phase transition as a function of the number of retained features. For physico-chemical descriptors, this transition is between a glass-like phase when the features are few and a liquid-like phase. The glass-like phase exhibits bimodal order-parameter distributions and Binder cumulant minima. In contrast, for structural descriptors the transition is less sharp. Remarkably, for physico-chemical descriptors the critical number of features identified from the DII coincides with the saturation of downstream binary classification performance. These results provide a principled, unsupervised criterion for minimal feature sets in protein classification and reveal distinct mechanisms of criticality across different feature types.

A new Real Fluid Quasi-Conservative (RFQC) finite volume method is developed to address the numerical simulation of real fluids involving shock waves in transcritical and phase-change flows. To eliminate the spurious pressure oscillations inherent in fully conservative schemes, we extend the classic five-equation quasi-conservative model, originally designed for two-phase flows, to real fluids governed by arbitrary equations of state (EoS). The RFQC method locally linearizes the real fluid EoS at each grid point and time step, constructing and evolving the frozen Grüneisen coefficient $Γ$ and the linearization remainder $E_0$ via two advection equations. At the end of each time step, the evolved $Γ$ and $E_0$ are utilized to reconstruct the oscillation-free pressure field, followed by a thermodynamic re-projection applied to the conserved variables. Theoretical analysis demonstrates that, in smooth regions, the energy conservation error of the RFQC method is a high-order term relative to the spatial reconstruction truncation error. In discontinuous regions, this error is determined by the entropy increase rate, thereby maintaining consistency with the inherent truncation error of shock-capturing methods. A series of numerical tests verifies that the method can robustly simulate complex flow processes with only minor energy conservation errors, including transcritical flows, phase transitions, and shock-interface interactions. The RFQC method is proven to be both accurate and robust in capturing shock waves and phase transitions.

As quantum computing transitions from theoretical physics to engineering applications, there is a growing need for accessible simulation tools that bridge the gap between abstract linear algebra and practical implementation. While text-based frameworks (like Qiskit or Cirq) are standard, they often present a steep learning curve for students and engineers accustomed to graphical system design. This paper introduces QuVI (Quantum Virtual Instrument), an open-source quantum circuit toolkit developed natively within the NI LabVIEW environment. Moving beyond initial proof-of-concept models, QuVI establishes a robust framework that leverages LabVIEW's "dataflow" paradigm-where wires represent data and nodes represent operations-to provide an intuitive, visual analog to standard quantum circuit notation while enabling the seamless integration of classical control structures like loops and conditionals. The toolkit's capabilities are demonstrated through the construction and visualization of fundamental quantum algorithms, verifying results against theoretical predictions. By translating "Block Diagrams" directly into quantum state evolutions ("Bloch Spheres"), QuVI offers educators and researchers a powerful platform for prototyping quantum logic without leaving the graphical engineering workspace.

Predicting favorable protein-peptide binding events remains a central challenge in biophysics, with continued uncertainty surrounding how nonlocal effects shape the global energy landscape. Here, we introduce peripheral surface information (PSI) entropy, a quantitative measure of the statistical variability in apolar and charged non-interacting surface (NIS) proportions across conformational ensembles. Using energy-directed molecular docking via HADDOCK3 and explicit-solvent molecular dynamics simulations, it is demonstrated that favorable binding partners exhibit emergent, low-entropy N-states (discrete macrostates in NIS state space) indicative of preferential apolar/charged surface configurations. Across dozens of peptides and multiple receptor systems (WW, PDZ, and MDM2 domains), dominant N-states persisted under varied docking parameters and initial conditions. An experimental meta-ensemble of WW domains from 36 high-resolution structures confirmed the presence of dominant NIS modes independent of in silico methodology, suggesting an evolutionary selection pressure toward specific NIS fingerprints. These findings establish PSI entropy as a thermoinformatic descriptor that encodes favorable binding constraints into unique statistical signatures of the NIS.

We present a method for parametrizing sub-grid processes in the Shallow Water equations. We define coarse variables and local spatial averages and use a feed-forward neural network to learn sub-grid fluxes. Our method results in a local parametrization that uses a four-point computational stencil, which has several advantages over globally coupled parametrizations. We demonstrate numerically that our method improves energy balance in long-term turbulent simulations and also accurately reproduces individual solutions. The neural network parametrization can be easily combined with flux limiting to reduce oscillations near shocks. More importantly, our method provides reliable parametrizations, even in dynamical regimes that are not included in the training data.

Studying the electrical signals generated by living cells is key to understanding numerous biological phenomena. Electrochromic optical recording (ECORE) uses the electrochromism exhibited by certain materials to noninvasively measure these signals in real time. In this work, we report on the development of ECORE based on a high-NA microscope objective. We demonstrate the recording of extracellular action potentials from cardiomyocytes with single-cell resolution and a high sensitivity of 3 μV, which compares favorably to the previous record for any ECORE setup. Combining ECORE with microscopy simplifies the optical setup, allows for the simultaneous imaging of specimens, and makes ECORE accessible to a broader community of researchers, allowing for a better understanding of the biological processes that are integral to life.

Accurate sensing of chemical concentrations is essential for numerous biological processes. The accuracy of this sensing, for small numbers of molecules, is limited by shot noise. Corresponding theoretical limits on sensing precision, as a function of sensing duration, have been well-studied in the context of quasi-static and randomly fluctuating concentrations. However, during development and in many other cases, concentration profiles are not random but exhibit predictable spatiotemporal patterns. We propose that leveraging prior knowledge of these structured profiles can improve and accelerate concentration sensing by utilizing information from current molecular binding events to predict future concentrations. By framing the constrained sensing problem as Bayesian inference over an allowed class of spatiotemporal profiles, we derive new theoretical limits on sensing accuracy. Our analysis reveals that maximum a posteriori (MAP) estimation can outperform the classical Berg-Purcell and maximum-likelihood (Poisson counting) limits, achieving a sensing precision of $δc/c = 1/\sqrt{a^2N}$, where $N$ is the number of binding events, and $a > 1$ in certain cases. Thus knowledge of the statistical structure of concentration profiles enhances sensing precision, providing a potential explanation for the rapid yet highly accurate cell fate decisions observed during development.

Population-level dynamics of social cohesion and its underlying mechanisms remain difficult to study. In this paper, we propose a network approach to measure the evolution of social cohesion at the population scale and identify mechanisms driving the change. We use twelve annual snapshots (2010-2021) of a population-scale social network from the Netherlands linking all residents through family, household, work, school, and neighbor relations. Results show that over this period, social cohesion, quantified as average closure in the network, declines by more than 15%. We demonstrate that the decline is not due to changes in demographic composition, but to rewiring in individual ego networks. Statistical models confirm a decreasing overlap of social contexts and greater geographical mobility as drivers. Residential relocation, however, temporarily increases closure, suggesting that local cohesion-seeking behavior can yield global network fragmentation, with implications for policies related to housing, urban planning, and social integration.

Perovskite solar cells (PSCs) have emerged as strong contenders for the next generation of photovoltaic (PV) technologies due to their exceptional light absorption properties, tunability, and affordability in manufacturing. Here, we presented an ingenious hexagonal nanowire (HNW)-based PSC that achieves broadband absorption, minimizes reflectance, and offers robust polarization insensitivity by improving light-matter interaction and increasing charge-collection efficiency. The rotational symmetry of the HNW configuration yielded polarization-independent absorbance under both TE and TM illumination across the visible and near-infrared spectra. The optimization of the geometrical parameters of CH3NH3PbI3-based HNW structure, including diameter, period, and fill ratio, offered a wide rangeof variations that influenced both optical properties and device performance. To further intensify photon confinement, a dielectric SiO2 sphere is partially embedded in the ITO layer, improving long-wavelength absorbance and increasing electron-hole pair generation near the active region. We analyzed the finite-difference time-domain (FDTD) method to examine the optical properties of our proposed structure. This study demonstrates that our proposed structure has achieved a higher generation rate, enhanced absorbance, and a higher optical short-circuit current density (Jsc) of 29.53 mA/cm2. Electrical performance is assessed by solving the coupled drift-diffusion and Poisson equations for the dynamics of carrier transport. The optimized HNW structure achieved a notable power conversion efficiency of 24.2%, highlighting a strong connection between optical confinement and effective carrier transport. These attributes render the proposed HNW PSC a viable option for high-performance PV systems and scalable thin-film solar technologies.

\texttt{aurel} is an open-source Python package designed to \emph{au}tomatically calculate \emph{rel}ativistic quantities. It uses an efficient, flexible and user-friendly caching and dependency-tracking system, ideal for managing the highly nonlinear nature of general relativity. The package supports both symbolic and numerical calculations. The symbolic part extends \texttt{SymPy} with additional tensorial calculations. The numerical part computes a wide range of tensorial quantities, such as curvature, matter kinematics and much more, directly from any spacetime and matter data arrays using finite-difference methods. Inputs can be either generated from analytical expressions or imported from Numerical Relativity (NR) simulations, with helper functions provided to read in data from standard NR codes. Given the increasing use of NR, \texttt{aurel} offers a timely post-processing tool to support the popularisation of this field.

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

Scaffolds have a key role in the clinical success of tissue engineering for the regeneration of damaged tissues. Their bio-performance is often described as the extent to which they can provide an extracellular matrix-like environment for cells embedded where their function and growth can effectively continue. For this purpose, tissue engineering scaffolds should exhibit biodegradability, biocompatibility, bioactivity, delivery, and mechanical performance. The use of polymer coatings, especially poly(lactic-co-glycolic acid) (PLGA), on tissue engineering scaffolds has been found to be one of the most effective methods to improve the scaffold properties. This paper reviews the techniques used to coat tissue engineering scaffolds with PLGA and its effects on the mechanical characteristics, biodegradability, biocompatibility, Molecular delivery, and osteointegration of the scaffolds. It is concluded that apart from apatite-formation ability, all bio-functionalities can be tuned through PLGA coatings. This reflects the great potential of this modification approach to be used in tissue regeneration and therapeutic delivery applications.

We present a comprehensive comparison of convolutional and transformer-based models for distinguishing quark and gluon jets using simulated jet images from Pythia 8. By encoding jet substructure into a three-channel representation of particle kinematics, we evaluate the performance of convolutional neural networks (CNNs), Vision Transformers (ViTs), and Swin Transformers (Swin-Tiny) under both supervised and self-supervised learning setups. Our results show that fine-tuning only the final two transformer blocks of the Swin-Tiny model achieves the best trade-off between efficiency and accuracy, reaching 81.4% accuracy and an AUC (area under the ROC curve) of 88.9%. Self-supervised pretraining with Momentum Contrast (MoCo) further enhances feature robustness and reduces the number of trainable parameters. These findings highlight the potential of hierarchical attention-based models for jet substructure studies and for domain transfer to real collision data.

Handwriting literacy is an important skill for learning and communication in school-age children. In the digital age, handwriting has been largely replaced by typing, leading to a decline in handwriting proficiency, particularly in non-alphabetic writing systems. Among children learning Chinese, a growing number have reported experiencing character amnesia: difficulty in correctly handwriting a character despite being able to recognize it. Given that there is currently no standardized diagnostic tool for assessing character amnesia in children, we developed an assessment to measure Chinese character amnesia in Mandarin-speaking school-age population. We utilised a large-scale handwriting dataset in which 40 children handwrote 800 characters from dictation prompts. Character amnesia and correct handwriting responses were analysed using a two-parameter Item Response Theory model. Four item-selection schemes were compared: random baseline, maximum discrimination, diverse difficulty, and an upper-and-lower-thirds discrimination score. Candidate item subsets were evaluated using out-of-sample prediction. Among these selection schemes, the upper-and-lower-thirds discrimination procedure yields a compact 30-item test that preserves individual-difference structure and generalizes to unseen test-takers (cross-validated mean r =.74 with full 800-item-test; within-sample r =.93). This short-form test provides a reliable and efficient tool of assessing Chinese character amnesia in children and can be used to identify early handwriting and orthographic learning difficulties, contributing to the early detection of developmental dysgraphia and related literacy challenges.

Zero-shot mutation prediction is vital for low-resource protein engineering, yet existing protein language models (PLMs) often yield statistically confident results that ignore fundamental biophysical constraints. Currently, selecting candidates for wet-lab validation relies on manual expert auditing of PLM outputs, a process that is inefficient, subjective, and highly dependent on domain expertise. To address this, we propose Rank-and-Reason (VenusRAR), a two-stage agentic framework to automate this workflow and maximize expected wet-lab fitness. In the Rank-Stage, a Computational Expert and Virtual Biologist aggregate a context-aware multi-modal ensemble, establishing a new Spearman correlation record of 0.551 (vs. 0.518) on ProteinGym. In the Reason-Stage, an agentic Expert Panel employs chain-of-thought reasoning to audit candidates against geometric and structural constraints, improving the Top-5 Hit Rate by up to 367% on ProteinGym-DMS99. The wet-lab validation on Cas12i3 nuclease further confirms the framework's efficacy, achieving a 46.7% positive rate and identifying two novel mutants with 4.23-fold and 5.05-fold activity improvements. Code and datasets are released on GitHub (https://github.com/ai4protein/VenusRAR/).

Coinfection phenomena are common in nature, yet there is a lack of analytical approaches for coinfection systems with a high number of circulating and interacting strains. In this paper, we investigated a coinfection SIS framework applied to N strains, co-circulating in a structured host population. Adopting a general formulation for fixed host classes, defined by arbitrary epidemiological traits such as class-specific transmission rates, susceptibilities, clearance rates, etc., our model can be easily applied in different frameworks: for example, when different host species share the same pathogen, in classes of vaccinated or non-vaccinated hosts, or even in classes of hosts defined by the number of contacts. Using the strain similarity assumption, we identify the fast and slow variables of the epidemiological dynamics on the host population, linking neutral and non-neutral strain dynamics, and deriving a global replicator equation. This global replicator equation allows to explicitly predict coexistence dynamics from mutual invasibility coefficients among strains. The derived global pairwise invasion fitness matrix contains explicit traces of the underlying host population structure, and of its entanglement with the strain interaction and trait landscape. Our work thus enables a more comprehensive study and efficient simulation of multi-strain dynamics in endemic ecosystems, paving the way to deeper understanding of global persistence and selection forces, jointly shaped by pathogen and host diversity.

Hyperkinetic movement disorders (HMDs) such as dystonia, tremor, chorea, myoclonus, and tics are disabling motor manifestations across childhood and adulthood. Their fluctuating, intermittent, and frequently co-occurring expressions hinder clinical recognition and longitudinal monitoring, which remain largely subjective and vulnerable to inter-rater variability. Objective and scalable methods to distinguish overlapping HMD phenotypes from routine clinical videos are still lacking. Here, we developed a pose-based machine-learning framework that converts standard outpatient videos into anatomically meaningful keypoint time series and computes kinematic descriptors spanning statistical, temporal, spectral, and higher-order irregularity-complexity features.

Antimicrobial resistance threatens healthcare sustainability and motivates low-cost computational discovery of antimicrobial peptides (AMPs). De novo peptide generation must optimize antimicrobial activity and safety through low predicted toxicity, but likelihood-trained generators do not enforce these goals explicitly. We introduce ProDCARL, a reinforcement-learning alignment framework that couples a diffusion-based protein generator (EvoDiff OA-DM 38M) with sequence property predictors for AMP activity and peptide toxicity. We fine-tune the diffusion prior on AMP sequences to obtain a domain-aware generator. Top-k policy-gradient updates use classifier-derived rewards plus entropy regularization and early stopping to preserve diversity and reduce reward hacking. In silico experiments show ProDCARL increases the mean predicted AMP score from 0.081 after fine-tuning to 0.178. The joint high-quality hit rate reaches 6.3\% with pAMP $>$0.7 and pTox $<$0.3. ProDCARL maintains high diversity, with $1-$mean pairwise identity equal to 0.929. Qualitative analyses with AlphaFold3 and ProtBERT embeddings suggest candidates show plausible AMP-like structural and semantic characteristics. ProDCARL serves as a candidate generator that narrows experimental search space, and experimental validation remains future work.

Predicting critical transitions in complex systems, such as epileptic seizures in the brain, represents a major challenge in scientific research. The high-dimensional characteristics and hidden critical signals further complicate early-warning tasks. This study proposes a novel early-warning framework that integrates manifold learning with stochastic dynamical system modeling. Through systematic comparison, six methods including diffusion maps (DM) are selected to construct low-dimensional representations. Based on these, a data-driven stochastic differential equation model is established to robustly estimate the probability evolution scoring function of the system. Building on this, a new Score Function (SF) indicator is defined by incorporating Schrödinger bridge theory to quantify the likelihood of significant state transitions in the system. Experiments demonstrate that this indicator exhibits higher sensitivity and robustness in epilepsy prediction, enables earlier identification of critical points, and clearly captures dynamic features across various stages before and after seizure onset. This work provides a systematic theoretical framework and practical methodology for extracting early-warning signals from high-dimensional data.

We investigate the effects of Hawking radiation on quantum steering and steering asymmetry in a tripartite system embedded in Schwarzschild spacetime. All tripartite steering types were classified,comprising three "1 to 2" and three "2 to 1" steering cases. Through a systematic analysis of all physically relevant scenarios (including accessible and inaccessible modes), we classify three canonical scenarios with one, two and three physically accessible modes. In the scenario of three physically accessible modes, Hawking radiation disrupts quantum steering, with the maximum steering asymmetry during the two-way steering to one-way steering transition precisely demarcating the phase boundary between these regimes. For two physically accessible modes, Hawking radiation exhibits dual behavior: enhancing the steering from Alice and Bob to anti-Charlie under certain parameters while suppressing it under others, while net strengthening other steering types. When considering one physically accessible mode, the Hawking effect of the black hole significantly enhances quantum steering. These findings provide new insights into quantum correlations in curved spacetime and establish observable signatures of Hawking effects in quantum steering phenomena.

We study spectral moments of the Bures-Hall random matrices ensemble. The main result establishes a recurrence relation for the $k$-th spectral moment valid for a real-valued $k$, in contrast to prevailing results in the literature of different ensembles of assuming an integer $k$. The key to establish the recurrence relation is the obtained Christoffel-Darboux formulas of correlation kernels of the ensemble that avoid tedious summations. As an application of our spectral moment results, we re-derive the formulas of average von Neumann entropy and quantum purity of Bures-Hall ensemble conjectured by Ayana Sarkar and Santosh Kumar. This work is dedicated to the memory of Santosh Kumar.

Superconducting qubits are a leading candidate for utility-scale quantum computing due to their fast gate speeds and steadily decreasing error rates. The requirement for millikelvin operating temperatures, however, creates a significant scaling bottleneck. Modular architectures using optical fiber links could bridge separate cryogenic nodes, but superconducting circuits do not have coherent optical transitions and microwave-to-optical conversion has not been shown for any non-classical photon state. In this work, we demonstrate the on-demand generation and tomographic reconstruction of itinerant single microwave photons at 8.9 GHz from a superconducting qubit. We upconvert this non-Gaussian state with a transducer added noise below 0.012 quanta and count the converted telecom photons at 193.4 THz with a signal-to-noise ratio of up to 5.1$\pm$1.1. We characterize the trade-offs between throughput and noise, and establish a viable path toward heralded entanglement distribution and gate teleportation. Looking ahead, these results empower existing superconducting devices to take a key role in distributed quantum technologies and heterogeneous quantum systems.

In quantum field theory, the algebraic existence of a field does not guarantee the existence of a corresponding localized asymptotic particle state. This distinction is well established in the presence of infrared effects, long-range correlations, and environmental interactions, and becomes particularly relevant in supersymmetric theories, where fermionic and bosonic degrees of freedom are constrained at the algebraic level but need not share identical asymptotic behavior. In this work we introduce a minimal and predynamical localization criterion that distinguishes algebraically allowed degrees of freedom from those capable of forming stable, phasecoherent asymptotic states. The criterion is formulated in terms of long-time stability under slow structural fluctuations of an effective background, without modifying the underlying field equations or introducing new physical interactions. We show that fermionic and scalar fields respond qualitatively differently to such structural effects. While fermionic modes may retain asymptotic stability, scalar modes generically exhibit decoherence and damping, preventing their interpretation as localized one-particle states. This provides a conservative and model-independent perspective on how supersymmetric algebraic structures may coexist with an asymmetric observable particle spectrum. The analysis is intentionally non-constructive and does not rely on specific supersymmetrybreaking mechanisms, cosmological assumptions, or new dynamical ingredients. Rather, it clarifies localization as an independent structural requirement for particle existence within standard quantum field theory.

One-sided device-independent quantum key distribution (1SDI-QKD) offers a practical middle ground between fully device-independent protocols and standard QKD, achieving security with detection efficiencies as low as 50.1\% on the untrusted side. However, prior analyses assumed idealized channels, neglecting realistic noise sources. We extend the 1SDI-QKD framework to include amplitude damping, dephasing, and depolarizing noise, quantifying their impact on secure key rates and efficiency requirements. Our results reveal a clear noise hierarchy: dephasing is most tolerable (secure keys achievable at 70\% efficiency with 30\% noise), while amplitude damping and depolarizing noise dramatically elevate requirements to over 90\%. Crucially, we find that security is lost while substantial entanglement remains (concurrence $C \approx 0.7$--$0.8$), demonstrating that steering violation, not merely entanglement, determines 1SDI-QKD security. To mitigate noise effects, we integrate the BBPSSW entanglement purification protocol, showing that 2--4 rounds can restore positive key rates in otherwise insecure regimes. Our resource overhead analysis reveals that effective key rates peak at moderate purification depths; excessive rounds become counterproductive. These findings establish practical boundaries for deploying 1SDI-QKD over metropolitan-scale quantum networks.

The standard $δN$ formalism, a cornerstone for calculating nonlinear curvature perturbations on super-Hubble scales, relies on the separate universe assumption, in which spatial gradients are neglected. However, this approximation breaks down in scenarios critical for primordial black hole formation, such as transitions to an ultra-slow-roll phase, where gradient interactions induce significant non-conservation of the comoving curvature perturbation. In this Letter, we introduce a framework that incorporates gradient corrections into the $δN$ formalism at a desired order by adding an effective source term to the background Klein--Gordon equation. This approach allows for a nonlinear treatment of curvature perturbations at the end of inflation considering initial conditions at the time of horizon exit. By computing the equilateral non-Gaussianity parameter $f_{\mathrm{NL}}^{\mathrm{eq}}$, we demonstrate that our method captures essential features missed by the standard $δN$, offering a simple yet rigorous pathway to determine nonlinear evolution expected from cosmological perturbation theory.

Symmetries play a central role in both equilibrium and nonequilibrium phase transitions, yet their quantitative characterization in dynamical quantum phase transitions (DQPTs) remains an open challenge. In this work, we establish a direct connection between symmetry properties of a many-body model and measures of quantum asymmetry, showing that asymmetry monotones provide a robust and physically transparent indicator of dynamical quantum criticality. Focusing on the quenched Lipkin-Meshkov-Glick model, we demonstrate that asymmetry measures associated with collective spin generators faithfully capture the onset of DQPTs, reflecting the dynamical restoration or breaking of underlying symmetries. Remarkably, the time-averaged asymmetry exhibits clear signatures of the dynamical critical point, in close correspondence with both the dynamical order parameter and the behavior of entropy production. We further uncover a quantitative link between asymmetry generation and thermodynamic irreversibility, showing that peaks in asymmetry coincide with maximal entropy production across the transition. Our results position asymmetry as a unifying concept bridging symmetry, information-theoretic quantifiers, and nonequilibrium thermodynamics in dynamical quantum phase transitions, providing a powerful framework for understanding critical dynamics beyond traditional order parameters.

Quantum dynamics retains a permanent and universal memory of its initial conditions, even in systems whose spectra display fully chaotic, random-matrix behavior. This effect, known as the quantum birthmark, appears as an enhancement of the long-time return probability of any non-stationary state compared to the overlap with a typical ergodic state. In this work, we develop the full theoretical foundation for this universal contribution that depends only on the global symmetry class and accessible Hilbert-space dimension, not on the microscopic dynamics. Our findings reveal that quantum evolution preserves an unavoidable, symmetry-controlled imprint of its origin, a quantum effect calling into question classical expectations of ergodicity and the resulting thermalization scenarios.

Strong theoretical arguments suggest that a breakdown of Lorentz Invariance could arise under some very particular conditions. From an experimental point of view, it is important to test the Local Lorentz Invariance with ever greater precision and in all contexts, regardless of the theoretical motivation for the possible violation. In this paper we discuss some aspects of the gravitational sector. Tests of Lorentz Invariance in the context of gravity are difficult and rare in the literature. Possible violations could arise from quantum physics applied to gravity or the presence of vector and tensor fields mediating the gravitational interaction together with the metric tensor of General Relativity. We present our results in the latter case. We analyzed the orbit of the LAGEOS and LAGEOS II satellites over a period of almost three decades. The effects of the possible preferred frame represented by the cosmic microwave background radiation on the mean argument of latitude of the satellites orbit were considered. These effects would manifest themselves mainly through the post-Newtonian parameter $α_1$, a parameter that has a null value in General Relativity. We constrain this parameterized post-Newtonian parameter down to the level of $α_1 \le 2\times10^{-5}$, improving a previous limit obtained through the Lunar Laser Ranging technique.

Bidirectional devices are devices for which the roles of the input and output ports can be exchanged. Mathematically, these devices are described by bistochastic quantum channels, namely completely positive linear maps that are both trace-preserving and identity-preserving. Recently, it has been shown that bidirectional quantum devices can, in principle, be used in ways that are incompatible with a definite input-output direction, giving rise to a new phenomenon called input-output indefiniteness. Here we characterize the most general forms of input-output indefiniteness, associated with a hierarchy of higher-order transformations built from transformations of bistochastic quantum channels. Some levels of the hierarchy correspond to transformations that combine bistochastic channels in a definite causal order, while generally using each channel in an indefinite input-output direction. For other levels of the hierarchy, the indefiniteness can involve both the local input-output direction of each process and the global causal order among the processes. On the foundational side, the hierarchy of higher-order transformations characterized here can be regarded as the largest set of physical processes compatible with a time-symmetric variant of quantum theory, where the possible state transformations are restricted to bistochastic channels.

The Planet Nine hypothesis encompasses a body of about 5-8 Earth's masses whose orbital plane would be inclined to the ecliptic by one or two tens of degrees and whose perihelion distance would be as large as about 240-385 astronomical units. Recently, a couple of his epigones have appeared: Planet X and Planet Y. The former is a sort of minor version of Planet Nine in that all its physical and orbital parameters would be smaller. Instead, the latter would have a mass ranging from that of Mercury to the Earth's one and semimajor axis within 100-200 astronomical units. By using realistic upper bounds for the orbital precessions of Saturn, one can get insights on their position which, for Planet Nine, appears approximately confined around its aphelion. Planet Y can be just a Mercury-sized object at no less than about 125 astronomical units, while Planet X appears to be ruled out. Dedicated data reductions by modeling such perturber(s) are required to check the present conclusions, to be intended as hints of what might be detectable should planetary ephemerides include them. A probe on the same route of Voyager 1 would be perturbed by Planet Nine by about 20-40 km after some decades.

Barren plateaus are typically characterized by vanishing gradients, yet the feasibility of curvature-based optimization fundamentally relies on the statistical resolvability of the Hessian matrix. In this work, we quantify the entrywise resolvability of the Hessian for Variational Quantum Algorithms at random initialization. By leveraging exact second-order parameter-shift rules, we derive a structural representation that reduces the variance of Hessian entries to a finite covariance quadratic form of shifted cost evaluations. This framework reveals two distinct scaling regimes that govern the sample complexity required to resolve Hessian entries against shot noise. For global objectives, we prove that Hessian variances are exponentially suppressed, implying that the number of measurement shots must scale as $O(e^{αn})$ with the number of qubits $n$ to maintain a constant signal-to-noise ratio. In contrast, for termwise local objectives in bounded-depth circuits, the variance decay is polynomial and explicitly controlled by the backward lightcone growth on the interaction graph, ensuring that curvature information remains statistically accessible with $O(\mathrm{poly}(n))$ shots. Extensive numerical experiments across varying system sizes and circuit depths demonstrate these theoretical bounds and the associated sampling costs. Our results provide a rigorous criterion for the computational tractability of second-order methods at initialization.

Energy Absorption Interferometry (EAI) is a technique for measuring the responsivities and complex-valued spatial polarimetric forms of the individual degrees of freedom through which a many-body system can absorb energy. It was originally formulated using the language of quantum correlation functions, making it applicable to different kinds of excitation (electromagnetic, elastic and acoustic fields). EAI has been applied in a variety of theoretical and experimental ways. It is particularly effective at characterising the multimode behaviour of ultra-low-noise far-infrared and optical devices, imaging arrays, and complete instruments, where it can be used to ensure that a system is maximally responsive to those partially coherent fields that carry signal whilst avoiding those that only carry noise. Despite its utility there is no comprehensive overview of electromagnetic EAI. In this paper we describe the theoretical foundations of the method, and present a range of new techniques in areas relating to sampling, phase referencing, mode reconstruction and noise. We present, for the first time, an analysis of how noise propagates through an experiment resulting in errors and artefacts on spectral and modal plots. A noise model is essential, because it determines the signal to noise ratio needed to ensure a given level of experimental fidelity.

In this paper, the quantum corrections to the kinematics of geometry, specifically geodesics, are presented. This is done by employing the path integral over the geodesics. Interestingly, the geodesics do not see any modifications in this framework. However for the distances, it is demonstrated that these quantum corrections exhibit distinct behaviors for time-like, light-like, and space-like geodesics. For time-like geodesics, the maximum correction is the Planck length, which disappears when the classical separation vanishes. The light-like geodesics do not exhibit quantum corrections, meaning that the causal light cone remains the same in both classical and quantum frameworks under certain conditions. The quantum corrections for space-like geodesics impose a minimum on space-like separation, potentially playing a role in removing singularities by preventing null congruences from being closer than the Planck length. This framework also explores the correspondence between space-like/time-like geodesics and quantum/statistical physics.

We describe a spin interferometer using ultracold YbF molecules and develop the complete set of techniques needed to measure the electron's electric dipole moment, $d_e$, with this apparatus. The molecules are cooled in an optical molasses and prepared in a single internal quantum state. A Raman transition prepares a spin superposition which evolves in parallel magnetic and electric fields before a second Raman transition maps the phase onto the populations of two hyperfine states. These populations are read out using detectors that have spatial and temporal resolution and approach unit efficiency. We characterize the efficiencies and fidelities of all these steps and evaluate the sensitivity of this approach to measuring $d_e$.

Quantum metrology harnesses quantum resources to achieve measurement precision beyond classical limits. This work investigates a Mach-Zehnder interferometer incorporating a Kerr nonlinear phase shifter, using photon-added two-mode squeezed coherent states generated via four-wave mixing as input. This study demonstrates that increasing both the photon-addition order and input resource strength systematically enhances phase sensitivity, quantum Fisher information, and the quantum Cramér-Rao bound. The system not only surpasses the standard quantum limit but also approaches or exceeds the Heisenberg limit with linear phase shifts, while Kerr nonlinearity enables overcoming the super-Heisenberg limit. The proposed scheme exhibits enhanced robustness against photon loss, providing a promising approach for high-precision quantum metrology applications.

Nitrogen-vacancy (NV) centers in diamond enable quantitative magnetic imaging, yet practical implementations must balance spatial resolution against acquisition time (and thus per-pixel sensitivity). Single-NV scanning magnetometry achieves genuine nanoscale resolution, nonetheless requires typically a slow pixel-by-pixel acquisition. Meanwhile, wide-field NV-ensemble microscopy provides parallel readout over a large field of view, however is jointly limited by the optical diffraction limit and the sensor-sample standoff. Here, we present a sparse-sampling strategy for reconstructing high-resolution wide-field images from only a small number of measurements. Using simulated NV-ensemble detection of ac magnetic fields, we show that a mean-adjusted Bayesian estimation (MABE) framework can reconstruct 10000-pixel images from only 25 sampling points, achieving SSIM values exceeding 0.999 for representative smooth field distributions, while optimized dynamical-decoupling pulse sequences yield an approximately twofold improvement in magnetic-field sensitivity. The method further clarifies how sampling patterns and sampling density affect reconstruction accuracy and suggests a route toward faster and more scalable magnetic-imaging architectures that may extend to point-scanning NV sensors and other magnetometry platforms, such as SQUIDs, Hall probes, and magnetic tunnel junctions.

We present a quantum circuit architecture for the one-to-two cloning of $N$-qubit registers. It implements the broad class of Pauli cloners by extending the Niu--Griffiths architecture to multi-qubit systems. In the single-qubit case, we provide explicit constructions for asymmetric universal, phase covariant and biased cloners. We explore the fundamental relationship between Pauli errors, mutually unbiased bases and Pauli cloning. Furthermore, we demonstrate how Pauli cloners can be tailored to specific noise models in the context of quantum communication, especially quantum key distribution.

We develop a deterministic zeroth-order mirror descent framework by replacing gradients with a general vector field, yielding a vector-field-driven mirror update that preserves Bregman geometry while accommodating derivative-free oracles. Our analysis provides a unified evaluation template for last-iterate function values under a relative-smoothness-type inequality, with an emphasis on trajectory-wise (a posteriori) certification: whenever a verifiable inequality holds along the realized iterates, we obtain explicit last-iterate guarantees. The framework subsumes a broad class of information-geometric algorithms, including generalized Blahut-Arimoto-type updates, by expressing their dynamics through suitable choices of the vector field. We then instantiate the theory with deterministic central finite differences in moderate dimension, where constructing the vector field via deterministic central finite differences requires 2d off-center function values (and one reusable center value), i.e., 2d+1 evaluations in total, where d is the number of input real numbers. In this deterministic finite-difference setting, the key interface property is not classical convexity alone but a punctured-neighborhood generalized star-convexity condition that isolates an explicit resolution-dependent error floor. Establishing this property for the finite-difference vector field reduces to a robust conic dominance design problem; we give an explicit scaling rule ensuring the required uniform dominance on a circular cone. Together, these results expose a hidden geometric structure linking Bregman telescoping identities, deterministic certification, and robust conic geometry in zeroth-order mirror descent.

Thermal light sources exhibiting photon bunching have been suggested for sensing applications that exploit timing correlations of stationary light, including range finding, clock synchronization, and non-line-of-sight imaging. However, these proposals have remained unrealized in practice because available sources of photon bunching either possess coherence times too short to be timing resolved by photodetectors, or produce brightness levels too low to tolerate realistic return losses. In this work, we demonstrate a low-loss method for generating photon bunching with a conversion efficiency nearly 9 orders of magnitude higher than that achieved by many other bunching processes.

We advance ``Latent entropy" (L-entropy) as a novel measure to characterize genuine multipartite entanglement in pure states, applicable to quantum systems with both finite and infinite degrees of freedom. This measure, derived from an upper bound on reflected entropy, attains its maximum for three-party GHZ states and $n=4,5$-party $2$-uniform states. We establish that it satisfies all essential properties of a genuine multipartite entanglement measure, including being a pure-state entanglement monotone. We further obtain an analogue of the Page curve by analyzing the behavior of L-entropy in multiboundary wormholes, emphasizing their connection to multipartite entanglement in random states. Specifically, for $n = 5$, we show that random states approximate $2$-uniform states, exhibiting maximal multipartite entanglement. Extending these ideas to finite temperatures, we introduce the Multipartite Thermal Pure Quantum (MTPQ) state, a generalization of the thermal pure quantum state to multipartite systems, and demonstrate that the entanglement structure in states of the multicopy SYK model exhibits finite-temperature $2$-uniform behavior.

Quantum reservoir computing (QRC) is a promising quantum machine learning framework for near-term quantum platforms, yet the performance of different QRC architectures under realistic constraints remains largely unexplored. Here, we provide a comparative numerical study of single-step-QRC (SS-QRC) and multi-step-QRC (MS-QRC) architectures implemented on a Rydberg atom array. We demonstrate that while MS-QRC performance is highly sensitive to the underlying dynamical phase of matter and decoherence, SS-QRC exhibits greater robustness. Using the randomized measurement toolbox to mitigate measurement overhead, we reveal that sampling noise undermines the convergence property required for MS-QRC. This leads to a significant reduction in the information processing capacity (IPC) of MS-QRC, deteriorating its performance on nonlinear time-series benchmarks. In contrast, SS-QRC maintains high IPC and accuracy across both temporal and non-temporal tasks. Our results suggest SS-QRC as a preferred candidate for near-term practical applications due to its resilience to system configurations and statistical noise.

Entropic uncertainty relations play a fundamental role in quantum information theory. However, determining optimal (tight) entropic uncertainty relations for general observables remains a formidable challenge and has so far been achieved only in a few special cases. Motivated by Schwonnek \emph{et al.} [PRL \textbf{119}, 170404 (2017)], we recast this task as a geometric optimization problem over the quantum probability space. This procedure leads to an effective outer-approximation method that yields tight entropic uncertainty bounds for general measurements in finite-dimensional quantum systems with preassigned numerical precision. We benchmark our approach against existing analytical and majorization-based bounds, and demonstrate its practical advantage through applications to quantum steering.

As demonstrated by observations, every stellar-mass object rotates around some axis; some objects spin faster than others due to different mechanisms. Furthermore, these spinning objects are slightly deformed and are no longer perfect spheres because of hydrostatic equilibrium. The well-known Kerr solution of the Einstein Field Equations (EFE) represents the spacetime surrounding a rotating spherical gravitational source. However, real objects deviate from a perfect sphere and may be prolate or oblate. There are several solutions of the EFE that represent the spacetime of deformed objects. The Kerr--like (KL) metric represents the spacetime surrounding this kind of object, where the deformation is characterized by the mass quadrupole moment parameter $q_{\mathrm{KL}}$. When $q_{\mathrm{KL}} \neq 0$, the Carter constant no longer exists and the equations of motion (EOM) are no longer integrable; therefore, the system exhibits chaotic orbits. Another widely used solution is the Hartle--Thorne (HT) metric, which has similar characteristics and represents a slightly deformed, slowly rotating star. The HT metric has several versions, and two of them were selected to test their validity. The traditional HT version, which contains logarithmic terms, is less accurate than the version with exponential terms. Moreover, both the KL and HT metrics may be extended to include contributions due to the magnetic dipole moment of the source. The equations of motion (EOM) were computed, and these new dynamical systems display several interesting features, which are shown in their Poincar'e sections.

We establish tight connections between entanglement entropy and the approximation error in Trotter-Suzuki product formulas for Hamiltonian simulation. Product formulas remain the workhorse of quantum simulation on near-term devices, yet standard error analyses yield worst-case bounds that can vastly overestimate the resources required for structured problems. For systems governed by geometrically local Hamiltonians with maximum entanglement entropy $S_\text{max}$ across all bipartitions, we prove that the first-order Trotter error scales as $\mathcal{O}(t^2 S_\text{max} \operatorname{polylog}(n)/r)$ rather than the worst-case $\mathcal{O}(t^2 n/r)$, where $n$ is the system size and $r$ is the number of Trotter steps. This yields improvements of $\tildeΩ(n^2)$ for one-dimensional area-law systems and $\tildeΩ(n^{3/2})$ for two-dimensional systems. We extend these bounds to higher-order Suzuki formulas, where the improvement factor involves $2^{pS^*/2}$ for the $p$-th order formula. We further establish a separation result demonstrating that volume-law entangled systems fundamentally require $\tildeΩ(n)$ more Trotter steps than area-law systems to achieve the same precision. This separation is tight up to logarithmic factors. Our analysis combines Lieb-Robinson bounds for locality, tensor network representations for entanglement structure, and novel commutator-entropy inequalities that bound the expectation value of nested commutators by the Schmidt rank of the state. These results have immediate applications to quantum chemistry, condensed matter simulation, and resource estimation for fault-tolerant quantum computing.

In this study, we investigate the possible existence of static and spherically symmetric wormhole solutions within the context of the newly formulated extended $f(Q,T)$ gravity. We analyze a linear model, $f(Q,T)=αQ+ βT$, and focus on traversable wormholes. By applying the variational method, we derive modified versions of the field equations that are influenced by an anisotropic matter source for a zero redshift function. It has been observed that the violation of energy conditions is influenced by the parameters $α$ and $β$. We reach the conclusion that solutions which violate the radial and lateral null energy condition in the context of general relativity may still adhere to the energy conditions within the realm of $f(Q,T)$ gravity. To begin with, by utilizing a linear equation of state for radial pressure, we obtain a power-law shape function. Additionally, we investigate solutions defined by a variable equation of state parameter. A broad spectrum of non-exotic wormhole solutions has been identified, contingent upon the particular parameters of the model.

We present a physics-informed classical-quantum machine learning framework for discriminating pristine CaF$_2$ from Er-doped CaF$_2$ using first-principles optical descriptors. Finite Ca$_8$F$_{16}$ and Ca$_7$ErF$_{16}$ clusters were constructed from the fluorite structure (a=5.46~$Å$) and treated using density functional theory (DFT) and linear-response time-dependent DFT (LR-TDDFT) within the GPAW code. Geometry optimization was performed in LCAO mode with a DZP basis and PBE exchange-correlation functional, followed by real-space finite-difference ground-state calculations with grid spacing h=0.30~$Å$ and N$_{bands}$=N$_{occ}$+20. Optical excitations up to 10~eV were obtained via the Casida formalism and converted into continuous absorption spectra using Gaussian broadening ($σ$=0.1-0.2~eV). From 1,589 energy-resolved points per system, physically interpretable descriptors including transition energy $E$, extinction coefficient $κ$, and absorption coefficient $α$ were extracted. A classical RBF-kernel support vector machine (SVM) achieves a test accuracy (ACC) of 0.983 and ROC-AUC of 0.999. Quantum support vector machines (QSVMs) evaluated on statevector and noisy simulators reach accuracies of 0.851 and 0.817, respectively, while execution on IBM quantum hardware yields a test-slice accuracy of 0.733 under finite-shot and decoherence constraints. A hybrid quantum neural network (QNN) with a 3-qubit feature map and depth-4 ansatz achieves a test accuracy of 0.93 and AUC of 0.96. Results here demonstrate that dopant-induced optical fingerprints form a robust, physically grounded feature space for benchmarking near-term quantum learning models against strong classical baselines.

We formulate the traversability of wormhole throats as a local and covariant null defocusing condition derived from the Raychaudhuri equation. Since unimodular gravity preserves the local geometric structure of spacetime, the null focusing properties of geodesic congruences are unchanged with respect to general relativity. We show that any genuinely traversable wormhole in unimodular gravity necessarily violates the null energy condition, establishing a local no--go theorem for wormholes supported by ordinary matter in this framework.

We investigate the superradiant instability of Ayón-Beato-García-de Sitter (ABG-dS) black holes under massless charged scalar perturbations using both time-domain evolutions and frequency-domain computations. We show that the instability occurs only for the spherically symmetric mode with $\ell=0$, whereas asymptotically flat ABG black holes remain stable in the massless limit, which underscores the essential role of the cosmological horizon in providing a confining boundary. We further study the dependence of the growth rate on the cosmological constant $Λ$, the scalar charge $q$, and the black hole charge $Q$, finding that it reaches a maximum at intermediate values of $Λ$ and $q$ and increases monotonically with $Q$. Compared with Reissner-Nordström-de Sitter black holes, ABG-dS black holes exhibit distinct instability characteristics due to the modified electrostatic potential induced by nonlinear electrodynamics.

The Ermakov Pinney equation and its associated invariant are shown to arise naturally in stationary quantum mechanics when the Schrodinger equation is expressed in Bohm Madelung form and the Hamiltonian is diagonal and separable. Under these conditions, the stationary continuity constraint induces a nonlinear amplitude equation of Ermakov Pinney type in each degree of freedom, revealing a hidden invariant structure that is independent of whether the evolution parameter is time or space. By reformulating the separated stationary equations in Sturm Liouville form and applying Liouville normalization, we demonstrate that the quantum potential is encoded as a curvature contribution of the self adjoint operator rather than appearing as an additional dynamical term. This correspondence preserves the standard probabilistic predictions of quantum mechanics while yielding exact stationary Bohmian amplitudes and their associated invariants. The resulting invariant-based formulation provides stationary guiding fields and clarifies the ontological status of Bohmian amplitudes as geometrically encoded structures rather than auxiliary dynamical additions. The results further show that stationary constrained Bohm Madelung systems naturally admit variational formulations whose extremals preserve the Ermakov Lewis invariant.

Recent observations from the DESI Collaboration indicate a preference for quintom dark energy, i.e., its equation of state evolves across the cosmological constant boundary $w=-1$. It is well known that models with single perfect fluid or single scalar field minimally coupled to Einstein gravity develop perturbative instabilities around the crossing, thereby cannot realize the quintom scenario. In this paper, we provide a method to circumvent the instability problem of these models by introducing the coupling of dark energy to the Nieh-Yan density in teleparallel gravity. We show that with this coupling the background evolution is not affected, but the dark energy perturbation is removed from the menu of dynamical degrees of freedom, thus avoiding the intrinsic difficulties in the old models.

Quantum phase transitions in many-body systems are fundamentally characterized by complex correlation structures, which pose computational challenges for conventional methods in large systems. To address this, we propose a hybrid quantum-classical attention model. This model uses an attention mechanism, realized through swap tests and a parameterized quantum circuit, to extract correlations within quantum states and perform ground-state classification. Benchmarked on the cluster-Ising model with system sizes of 9 and 15 qubits, the model achieves high classification accuracy with less than 100 training data and demonstrates robustness against variations in the training set. Further analysis reveals that the model successfully captures phase-sensitive features and characteristic physical length scales, offering a scalable and data-efficient approach for quantum phase recognition in complex many-body systems.

In open quantum systems, reduced dynamics is commonly described by a master equation, whose Liouvillian gap closing (LGC) typically signals the emergence of decoherence-free subspace. By contrast, the dynamics of the full system-environment compound is governed by the underlying Hamiltonian spectrum, where bound states in the continuum (BICs) can protect long-lived quantum resources. Despite these parallel perspectives, the relation between LGC and BIC formation has remained largely unexplored. Here we bridge this gap in a paradigmatic giant-atom waveguide platform and show that the occurrence of LGC necessarily benchmarks the presence of a BIC in the full Hamiltonian description. By engineering the giant-atom geometry, we further demonstrate rich dynamical regimes-including Rabi oscillations, fractional decay, and complete exponential relaxation-depending on the number of supported BICs, which can be tuned from three to zero. Remarkably, when two BICs become frequency-degenerate, the long-time dynamics approaches a steady state rather than exhibiting persistent oscillations. Our results establish a direct spectral-dynamical connection between effective Markovian and underlying non-Markovian descriptions, and provide a route toward flexible control of open-system dynamics.

Can superconducting order be made a robust attractor of open-system dynamics in strongly correlated lattices? We demonstrate that it can by proposing a minimal dissipation-engineering protocol for the particle--hole symmetric Hubbard model. By applying a rotated quantum jump operator, a locally transformed $η$-pair lowering operator, on as little as a single lattice site, we show that the Lindblad evolution autonomously pumps the system from the vacuum into a nonequilibrium steady state (NESS) with macroscopic $η$-pair off-diagonal long-range order (ODLRO). Crucially, this local-to-global synchronization contrasts with schemes requiring spatially extensive reservoirs: here, a strictly local dissipative seed suffices to establish coherence across the interacting system. We elucidate the mechanism via local dark-state selection, controlled elimination of off-manifold excursions induced by hopping, and a Liouvillian invariant-subspace structure that yields an attractive fixed point with a finite dissipative gap. Furthermore, we classify the stability of this NESS against static disorder, identifying a broad regime where the superconducting attractor is resilient to Hamiltonian perturbations that leave the effective subspace structure intact, while pinpointing specific perturbations that directly dephase the $η$-pseudospin coherence and suppress ODLRO. Our results establish a disorder-tolerant route to stabilizing superconducting order as a non-thermal attractor via minimal local quantum-jump control.

Three decades ago, Ted Jacobson surprised us with a very appealing approach to classical gravity. According to his proposal, the gravitational field equations are the consequence of the first law of thermodynamics applied, locally, to a Rindler observer. Together with the dynamical laws of black holes, Jacobson's approach has become a very strong piece of evidence supporting the intimate connection between gravity, thermodynamics, and quantum theory. Jacobson's approach being formulated for Riemannian geometries, we have wondered what its consequences would be for non-Riemannian geometries, i.e., those that involve torsion, non metricity, or both. The results of our quest have been particularly appealing: we have found that the gravitational theory that derives from the Einstein-Hilbert action, arguably ``the simplest one'', does not belong to the pool of gravitational theories available for Nature's selection (except in the Riemannian case). In the search of a unique alternative, we have also considered the hypotheses employed in the formulation of the Lanczos-Lovelock theories of gravity. Together, the two approaches point towards the gravitational theory that derives from the Einstein-Hilbert action plus a term quadratic in the torsion vector as the one that would be selected by Nature in the non-Riemannian case without non metricity. The same strategy cannot be followed in the full non-Riemannian case as the two approaches are mutually inconsistent.

We embed the covariant, gauge-invariant gravitational radiation criteria of Fernández-Álvarez and Senovilla, based in terms of conformal geometry and the Bel-Robinson tensor, into the hydrodynamic framework of gauge/gravity duality. This construction uncovers a direct correspondence between bulk gravitational waves and dissipative processes in the boundary theory, from which a natural notion of entropy production emerges. We further analyze a smooth flat limit in which the dual fluid becomes Carrollian, with dissipation governed by Carroll-covariant tensors. As an example, we apply our framework to the Robinson-Trautman family of solutions.

Quantum chaos in isolated quantum systems is intimately linked to thermalization and the rapid relaxation of observables. Although the spectral properties of the chaotic phase in the tilted Bose-Hubbard model have been well characterized, the corresponding dynamical signatures across the transition to regularity remain less explored . In this work, we investigate this transition by analyzing the time evolution of the survival probability, the single-site entanglement entropy, and the half-chain imbalance. Our results reveal a clear hierarchy in the sensitivity of these observables: the relaxation value of the entanglement entropy varies smoothly as a function of the Hamiltonian parameters across the chaos-regular transition, while the imbalance exhibits a more pronounced distinction. Most notably, the survival probability emerges as the most robust indicator of the transition between chaos and regularity. When appropriately scaled, all three observables converge onto a common behavior as a function of the Hamiltonian parameters for different numbers of sites and bosons,enabling a universal characterization of the transition between chaotic and regular dynamics.

This thesis addresses a fundamental problem in deformation quantization: the difficulty of calculating the star-exponential, the symbol of the evolution operator, due to convergence issues. Inspired by the formalism that connects the star-exponential with the quantum propagator for bosonic systems, this work develops the analogous extension for the fermionic case. A rigorous method, based on Grassmann variables and coherent states, is constructed to obtain a closed-form expression for the fermionic star-exponential from its associated propagator. As a primary application, a fermionic version of the Feynman-Kac formula is derived within this formalism, allowing for the calculation of the ground state energy directly in phase space. Finally, the method is validated by successfully applying it to the simple and driven harmonic oscillators, where it is demonstrated that a simplified ("naive") approach (with an ad-hoc "remediation") is a valid weak-coupling limit of the rigorous ("meticulous") formalism, thereby providing a new and powerful computational tool for the study of fermionic systems.

Large-momentum-transfer techniques are instrumental for the next generation of atom interferometers as they significantly improve their sensitivity. State-of-the-art implementations rely on elastic scattering processes from optical lattices such as Bloch oscillations or sequential Bragg diffraction, but their performance is constrained by imperfect pulse efficiencies. Here we develop a Floquet-based theoretical framework that provides a unified description of elastic light-atom scattering across all relevant regimes. Within this formalism, we identify practical regimes that exhibit orders of magnitude reduced losses and improved phase accuracy compared to previous implementations. The model's validity is established through direct comparison with exact numerical solutions of the Schrödinger equation and through quantitative agreement with recent experimental benchmark results. These findings delineate previously unexplored operating regimes for large-momentum-transfer beam splitters and open new perspectives for precision atom-interferometric measurements in fundamental physics, gravity gradiometry or gravitational wave detection.

Quantum kernel methods offer significant theoretical benefits by rendering classically inseparable features separable in quantum space. Yet, the practical application of Quantum Machine Learning (QML), currently constrained by the limitations of Noisy Intermediate-Scale Quantum (NISQ) hardware, necessitates effective strategies to compress and embed large-scale real-world data like images into the constrained capacities of existing quantum devices or simulators. To this end, we propose Quantum Generator Kernels (QGKs), a generator-based approach to quantum kernels, comprising a set of Variational Generator Groups (VGGs) that merge universal generators into a parameterizable operator, ensuring scalable coverage of the available quantum space. Thereby, we address shortcomings of current leading strategies employing hybrid architectures, which might prevent exploiting quantum computing's full potential due to fixed intermediate embedding processes. To optimize the kernel alignment to the target domain, we train a weight vector to parameterize the projection of the VGGs in the current data context. Our empirical results demonstrate superior projection and classification capabilities of the QGK compared to state-of-the-art quantum and classical kernel approaches and show its potential to serve as a versatile framework for various QML applications.

Loops of virtual particles from the vacuum of quantum field theory (QFT) render black holes tidally deformable. We compute the static tidal response of unspinning charged black holes at arbitrary radius, using the perturbative formalism developed in 2501.18684. Since the gravitational and electromagnetic tidal responses mix, we generalize the notion of Love numbers to Love matrices. We derive the coupled equations of motion for the metric and electromagnetic fluctuations around purely electric and magnetic backgrounds. For large charged black holes, which are described by the Effective Field Theory (EFT) of gravity, we compute the full set of Love matrices induced by an arbitrary tower of $F^{2n}$ operators. We find that, although quantum corrections break electromagnetic duality, the Love matrices in electric and magnetic backgrounds are related by a $Z_2$ symmetry under electric-magnetic exchange. Going beyond EFT, we compute the Love matrices of small magnetic black holes. We show that the running of the Love matrices is governed by the running of the $U(1)$ gauge coupling, and we derive the correspondence between Love and $U(1)$ beta functions for arbitrary harmonics. The overall picture that emerges is that the QFT-induced tidal response of magnetic black holes saturates in the strong-field regime. These results imply that nearly-extremal magnetic black holes charged under an Abelian dark sector could be probed by gravitational-wave observations.

We study Weyl-invariant purely gravitational theories formulated within the Einstein-Cartan framework. In the Einstein-frame description, these models are dynamically equivalent to standard general relativity coupled to an axion-like pseudoscalar degree of freedom, which naturally drives a period of cosmic inflation. Without committing to a specific microscopic mechanism for reheating, we demonstrate that the post-inflationary reheating dynamics play a crucial role in shaping the inflationary predictions. In particular, we show that assumptions about the reheating temperature and the equation-of-state parameter can significantly affect the predicted values of inflationary observables, highlighting the necessity of consistently incorporating reheating effects in the phenomenological analysis of inflationary models.

A simple model is proposed to explain the recently found wave-like enhancement of the $Λ\barΛ$ pair production near the threshold at the $e^-e^+$ annihilation, which allows extracting parameters of the $Λ\barΛ$ interaction and the $Λ$-hyperon rms-radius from the oscillatory nature of the measured cross section. In particular, it predicts a single bound state of $Λ\barΛ$ with a binding energy of $\varepsilon_{Λ\barΛ}=(36\pm5)$MeV. The model is a generalization of the formulas obtained in our earlier work [1] to explain the effect of wave-like amplification found in it near the threshold of fusion reactions screened by a muon or electron. The analysis allows us to conclude that the effect of wave-like amplification is an integral feature of any two-particle near-threshold reaction. In this regard, it seems promising to investigate, within the framework of our model, the oscillatory nature of the electromagnetic form factors of hyperons and nucleons extracted in experiments on $e^- e^+$ annihilation. A natural further development of the model could be its generalization to processes of producing various hadron pairs in $e^-e^+$ annihilation.

Correlated microwave photon sources are key enablers for technologies in quantum-limited sensing, signal amplification and communication, but the reliance on millikelvin operating temperature limits their scalability for broader applications. Here, at room temperature, we demonstrate strong correlated microwave signals emitted from a hybrid magnon-photon platform. Different from traditional parametrically induced magnons with degenerate frequencies, we achieve non-degenerate excitations by coupling magnon modes simultaneously with two cavity photon modes. Through the magnon-photon interactions in the corresponding linear and nonlinear regimes, one input microwave photon splits into a pair of magnon polaritons that possess distinct frequencies but maintain strong inter-mode correlations. The nonlinear magnon polariton dynamics empowered by this new parametric platform brings both verified true randomness and robust multi-channel correlations, from which we construct a microwave communication experiment for noise resilient signal transmission with added security. This work establishes cavity magnonics as a versatile and compact platform for generating correlated multi-mode microwave signals, opening new avenues for applications in classical and quantum domains.

The Dirac-Bergmann algorithm for the Hamiltonian analysis of constrained systems is a nice and powerful tool, widely used for quantization and non-perturbative counting of degrees of freedom. However, certain aspects of its application to systems with first-class constraints are often overlooked in the literature, which is unfortunate, as a naive treatment leads to incorrect results. In particular, when transitioning from the total to the extended Hamiltonian, the physical information encoded in the constrained modes is lost unless a suitable redefinition of gauge invariant quantities is made. An example of this is electrodynamics, in which the electric field gets an additional contribution to its longitudinal component in the form of the gradient of an arbitrary Lagrange multiplier. Moreover, Dirac's conjecture, the common claim that all first-class constraints are independent generators of gauge transformations, is somewhat misleading in the standard notion of gauge symmetry used in field theories. At the level of the total Hamiltonian, the true gauge generator is a specific combination of primary and secondary first-class constraints; in general, Dirac's conjecture holds only in the case of the extended Hamiltonian. The aim of the paper is primarily pedagogical. We review these issues, providing examples and general arguments. Also, we show that the aforementioned redefinition of gauge invariants within the extended Hamiltonian approach is equivalent to a form of the Stueckelberg trick applied to variables that are second-class with respect to the primary constraints.

Understanding entanglement cost in non-local quantum computation (NLQC) is relevant to complexity, cryptography, gravity, and other areas. This entanglement cost is largely uncharacterized; previous lower bound techniques apply to narrowly defined cases, and proving lower bounds on most simple unitaries has remained open. Here, we give two new lower bound techniques that can be evaluated for any unitary, based on their controllable correlation and controllable entanglement. For Haar random two qubit unitaries, our techniques typically lead to non-trivial lower bounds. Further, we obtain lower bounds on most of the commonly studied two qubit quantum gates, including CNOT, DCNOT, $\sqrt{\text{SWAP}}$, and the XX interaction, none of which previously had known lower bounds. For the CNOT gate, one of our techniques gives a tight lower bound, fully resolving its entanglement cost. The resulting lower bounds have parallel repetition properties, and apply in the noisy setting.

Compatibility with existing classical network infrastructure offers a scalable path towards deploying largescale quantum networks. Here, we demonstrate O-band polarization-encoded quantum entanglement distribution over an installed 24.4-km fiber while coexisting with a state-of-the-art fully-loaded C-band classical communications line system and a picosecond-level precision L-band synchronization signal. The classical system carries two 800-Gbps channels while the remainder of the C-band is filled with amplified spontaneous emission as is standard for such state-of-the-art communications systems. We examine the spontaneous Raman scattering spectrum generated from this broadband C-band light and offer insights into wavelength allocation for O-band quantum channels. Optimal wavelength selection and narrow filtering enable well-preserved Bell state fidelity when coexisting with 21.4-dBm aggregate launch power across the C-band suitable for 36 Tbps transmission. To the best of our knowledge, this is the first implementation of entanglement-based quantum communications between two remote nodes coexisting with independent classical communications traffic. We demonstrate coexistence of quantum entanglement with ultra-high power levels and record classical bandwidth, offering promise for real-world entanglement-based networking integrated within high-capacity communications infrastructure.

We propose and analyze a versatile and efficient multiparameter quantum sensing protocol, which simultaneously estimates many non-commuting and time-dependent signals that are coherently or incoherently coupled to sensing particles. Even in the presence of control imperfections and readout errors, our approach can detect exponentially many parameters in the system size while maintaining the optimal scaling of sensitivity. To accomplish this, scrambling dynamics are leveraged to map distinct signals to unique patterns of bitstring measurements, which distinguishes a large number of signals without significant sensitivity loss. Based on this principle, we develop a computationally efficient protocol utilizing random global Clifford unitaries and evaluate its performance both analytically and numerically. Our protocol naturally extends to scrambling dynamics generated by random local Clifford circuits, local random unitary circuits (RUCs), and ergodic Hamiltonian evolution--commonly realized in near-term quantum hardware--and opens the door to applications ranging from precise noise benchmarking of quantum dynamics to learning time-dependent Hamiltonians.

Photon-graviton conversion in an electromagnetic field is a well-known prediction of Einstein-Maxwell theory. First discussed at tree-level by Gertsenshtein in 1962, more recently it has been shown to lead to magnetic dichroism starting from one-loop. While previously only two diagrams were assumed to contribute to this one-loop photon-graviton amplitude in a constant electromagnetic field, here we point out the existence of a third one involving a tadpole subdiagram. As shown by H. Gies and one of the authors in 2016 for the pure QED case, such diagrams cannot be omitted in general even though the tadpole formally vanishes. After a short review of the calculation of one-loop photon-graviton amplitudes in the worldline formalism, we use this formalism for a unified calculation of all three diagrams. Although phenomenologically this amplitude is mainly of interest for the case of the spinor loop in a magnetic field, here we will also include the scalar loop and the electric field component, since the computational effort is essentially the same. We show that the tadpole diagram, although contributing to the amplitude, does not contribute to the magnetic dichroism. The gravitational Ward identity provides a useful check.

Superconducting nanowire single-photon detectors are central to applications across quantum information science. Yet, their performance is limited by the effects of disorder and electrodynamic inhomogeneities that are not well understood. By combining DC transport, dark-count measurements, and bias-dependent microwave transmission spectroscopy in the presence of controlled nanoscale disorder introduced through helium-ion irradiation, we distinguish local instability-driven processes from intrinsic superconducting depairing and kinetic inductance nonlinearities. This approach enables systematic tuning of kinetic inductance, depairing currents, microwave dissipation, and mode structure within a single device. Bias- and temperature-dependent resonance shifts quantify disorder-induced modifications of the superconducting density of states through the nonlinear kinetic inductance, while the emergence of multiple resonant modes reveals the formation of electrodynamically distinct superconducting regions. Comparing depairing under current, field, and temperature isolates the dominant microwave loss mechanisms, separating vortex, quasiparticle, and two-level-system contributions, thus providing a robust multifunctional foundation for disorder engineering of superconducting nanowire detectors and resonators.

We study theoretically how high-gain effects affect the measurement outcome of visible signal spectra in undetected photon measurement schemes. We consider two interferometric configurations: firstly, the SU(1,1) interferometer where the idler incurs loss and additional dispersion in between two identical, lossless, squeezers; secondly, the induced coherence interferometer where the idler incurs loss and additional dispersion in between two identical, lossless, squeezers and where the second squeezer is seeded by the idler and a vacuum ancilla mode. Furthermore, we consider a distributed loss configuration where the idler incurs loss as it propagates in the nonlinear medium. Motivated by experimental evidence and due to the fact that broadband sources are ideal for these measurement schemes, we use the dispersive data of a third-order dispersion engineered integrated waveguide parametric down conversion (PDC) source presented in New Journal of Physics 26, 123025 (2024) to model the PDC spectra in the three configurations. For each configuration we consider the case of idler-only (i) absorption, (ii) additional dispersion, and (iii) the combined effects. We obtain results which outline the strength and weaknesses of the different configurations at different operation points.

In our paper we consider the validity of slow-roll approximations for Gauss-Bonnet inflation introduced in [1]. In contrast to the cited paper where the coupling function before the Gauss-Bonnet term have been chosen as a decaying function of the scalar field, here we consider growing coupling functions. We have found that while in [1] new slow-roll approximations work considerably better, now they do not increase the precision. Moreover, we identify some cases where more involved approximations work worse than the standard one. Corresponding explanations of such a situation are given.

In quantum physics, multiparticle systems are described by quantum states acting on tensor products of Hilbert spaces. This product structure leads to the distinction between product states and entangled states; moreover, one can quantify entanglement by considering the distance of a quantum state to the set of product states. The underlying optimization problem occurs frequently in physics and beyond, for instance in the computation of the injective tensor norm in multilinear algebra. Here, we introduce a method to determine the maximal overlap of a pure multiparticle quantum state with product states based on considering several copies of the pure state. This leads to three types of hierarchical approximations to the problem, all of which we prove to converge to the actual value. Besides allowing for the computation of the geometric measure of entanglement, our results can be used to tackle optimizations over stochastic local transformations, to find entanglement witnesses for weakly entangled bipartite states, and to design strong separability tests for mixed multiparticle states. Finally, our approach sheds light on the complexity of separability tests.

Gravitational wave (GW) astronomy opens a new venue to explore the universe. Future observatories such as LISA, the Laser Interferometer Space Antenna, are expected to observe previously undetectable fundamental physics effects in signals predicted by General Relativity (GR).One particularly interesting such signal is associated to the displacement memory effect, which corresponds to a permanent deformation of spacetime due to the passage of gravitational radiation. In this work, we explore the ability of LISA to observe and characterize this effect. In order to do this, we use state-of-the-art simulations of the LISA instrument, and we perform a Bayesian analysis to assess the detectability and establish general conditions to claim detection of the displacement memory effect from individual massive black hole binary (MBHB) merger events in LISA. We perform parameter estimation both to explore the impact of the displacement memory effect and to reconstruct its amplitude. We discuss the precision at which such a reconstruction can be obtained thus opening the way to tests of GR and alternative theories. To provide astrophysical context, we apply our analysis to black hole binary populations models and estimate the rates at which the displacement memory effect could be observed within the LISA planned lifetime.

We introduce a new causal spinfoam vertex for $4$d Lorentzian quantum gravity. The causal data are encoded in Toller $T$-matrices, which add to Wigner $D$-matrices $T^{(+)}+T^{(-)}=D$, and for which we provide a Feynman $\mathrm{i}\varepsilon$ representation. We discuss how the Toller poles cancel in the EPRL vertex, how the Livine-Oriti model is obtained in the Barrett-Crane limit, and how spinfoam causal data are distinct from Regge causal data. In the large-spin limit, we show that only Lorentzian Regge geometries with causal data compatible with the spinfoam data are selected, resulting in a single exponential $\exp(+\mathrm{i}\, S_{\mathrm{Regge}}/\hbar)$ and a new form of causal rigidity.

We propose a superconducting-circuit architecture that realizes a compact U(1) lattice gauge theory using the intrinsic infinite-dimensional Hilbert space of phase and charge variables. The gauge and matter fields are encoded directly in the degrees of freedom of the rotor variables associated with the circuit nodes, and Gauss's law emerges exactly from the conservation of local charge, without auxiliary stabilizers, penalty terms, or Hilbert-space truncation. A minimal gauge-matter coupling arises microscopically from Josephson nonlinearities, whereas the magnetic plaquette interaction is generated perturbatively via virtual matter excitations. Numerical diagonalization confirms the emergence of compact electrodynamics and coherent vortex excitations, underscoring the need for large local Hilbert spaces in the continuum regime. The required circuit parameters are within the current experimental capabilities. Our results establish superconducting circuits as a scalable, continuous-variable platform for analog quantum simulation of non-perturbative gauge dynamics.

Motivated by Maldacena's observer-centric formulation of de~Sitter physics \cite{Maldacena:2024spf}, we develop an observer-dependent state-counting framework in Euclidean de~Sitter space by modeling the observer as a massive equatorial worldline carrying an SU(3) clock. Starting from the gauge-fixed graviton path integral on $S^D$, we trace the one-loop phase $\ii^{D+2}$ to a finite set of scalar and conformal Killing modes and show that, once the worldline is included, the $(D-1)$ transverse negative modes cancel the corresponding $(D-1)$ conformal Killing directions mode by mode. The residual fixed-$β$ phase from the global conformal factor and reparametrizations is removed by imposing the Hamiltonian constraint $H_{\text{patch}} - H_{\text{clock}} - ν= 0$ via a Bromwich inverse Laplace transform, which under explicit complete-monotonicity assumptions yields a real and positive microcanonical density. We stress that this positivity statement is conditional on Assumptions (A1)--(A3) and is established at one loop about the round $S^D$ saddle in the probe regime $G E_{\rm clock}/R\ll 1$; a self-consistent backreacting or higher-loop extension is a natural next step. In earlier work \cite{Ali:2025wld,Ali:2024rnw} we argued that unbroken SU(3) confinement at $T\to 0$ can account for the observed value of the cosmological constant and for the origin of the fundamental constants $(\hbar,G,c)$ as effective couplings fixed by the SU(3) vacuum structure; this makes SU(3) the natural candidate for the internal clock of de~Sitter, whose radius and temperature are themselves set by the same cosmological constant. This idea is implemented with three explicit SU(3) realizations (qutrit, Cartan weight-lattice, and $U(1)^2$ rotor), for which the observer-inclusive density of states factorizes into a universal gravity factor, a universal worldline residue, and a clock-dependent SU(3) weight.

Pairs of entangled photons are crucial for photonic quantum technologies. The demand for integrability and multi-functionality suggests 'flat' platforms - ultrathin layers and metasurfaces - as sources of photon pairs. With the success in demonstrating spontaneous parametric down-conversion (SPDC) from such sources, an alternative process to generate photon pairs, spontaneous four-wave mixing (SFWM), also starts to attract interest. In materials with nonzero second-order nonlinear susceptibility $χ^{(2)}$, SFWM can generate photon pairs both directly, through the third-order nonlinear susceptibility $χ^{(3)}$, and in a cascaded way, through second harmonic generation (SHG) followed by SPDC. Usually, the cascaded process is more efficient. Here, we show that in a thin layer, direct SFWM dominates, because the wavevector mismatch for SFWM is much smaller than for SHG or SPDC. To demonstrate it, we implement the photon pair generation via SFWM in a second-order nonlinear material - a thin layer of lithium niobate (LN). The existence of both second- and third-order nonlinear processes offers broader opportunities for quantum state engineering.

A perfect d-dimensional quantum channel can convey log d-bits of classical information by encoding messages in d-orthogonal quantum states. Alternatively, for every quantum state at the senders end, there exist d-encoding operations which produce d-orthogonal quantum states. Transmitting which via a d-level perfect quantum channel it is possible to communicate log d-bits of classical information. But what if the set of encoding operations is restricted only within a physically constrained class? Here, we consider such a class of encoding operations to be the set of free operations for any quantum resource theory and show that the constrained capacity - namely, the free encoding capacity (FEC) emerged as a unit of the corresponding quantum resource. Moreover, we show that for the pointed resource theories - a resource theory admitting only a single free state - FEC becomes a faithful resource measure also. We also discuss the implications of FEC in the question of resource-theoretic state transformations and the possibility of extending its faithfulness for general quantum resource theories.

Fault-tolerant quantum computing with surface codes can be achieved by compiling logical circuits into lattice-surgery instructions. To minimize space-time volume, we present TopoLS, a topological compiler that combines ZX-diagram optimizations with Monte Carlo tree search guided by different operation placements and topology-aware circuit partitioning. Our approach enables scalable exploration of lattice surgery structures and consistently reduces resource overhead. Evaluations of various benchmark algorithms across multiple architectures show that TopoLS achieves an average 33% reduction in space-time volume over prior heuristic-based compilers, while maintaining linear compilation time scaling. Compared to the SAT-solver-based compiler, which provides optimal results only for small circuits before becoming intractable, TopoLS offers an effective and scalable solution for lattice-surgery compilation.

The generation of accurate waveforms from binary black hole (BBH) mergers is a major effort in Gravitational-Wave Astronomy. In recent years, machine-learning-based surrogate models for BBH waveforms have been proposed. Those offer the potential to dramatically accelerate waveform generation while maintaining accuracy competitive with that of traditional waveform approximants. In this work, we investigate the viability of autoencoders as generative models for gravitational-wave signals from quasi-circular BBH mergers. We introduce AESur3dq8, a novel surrogate waveform model based on autoencoders that enables the rapid and accurate construction of large template banks, producing millions of waveforms in under a second using modest computational resources. The model is trained on the numerical-relativity-informed surrogate NRHybSur3dq8 and subsequently fine-tuned using the SXS catalog of BBH simulations. We demonstrate that waveforms generated by AESur3dq8 achieve mismatches of order $10^{-4}$ with respect to Numerical Relativity waveforms, and that parameter estimation performed with these templates yields results fully consistent with those reported by the LIGO-Virgo-KAGRA Collaboration for observed gravitational-wave events.

Inferring the intrinsic population of compact binary mergers is complicated by detector selection biases and measurement uncertainties. Traditional parametric methods are limited by the need to presuppose functional forms, introducing model-dependent biases. To overcome these limitations, we introduce an inference framework powered by deep generative modeling. We develop a flexible, data-driven population model using a Correlated Compound-Mixture Density Network. This architecture integrates mixture models to handle multimodality, Gaussian copulas for parameter dependencies, and a library of flexible marginal distributions. The network is trained to approximate the posterior distribution of the population's hyperparameters using amortized variational inference with Normalizing Flows on catalogs of gravitational-wave events. We demonstrate the framework's capabilities in two distinct regimes. First, using simulated catalogs of supermassive black hole binary mergers for the Laser Interferometer Space Antenna (LISA), we show that the method accurately recovers complex three-dimensional distributions and absolute merger rates from sparse datasets, effectively correcting for selection effects and measurement uncertainties. Second, we validate the framework on real observational data from the LIGO-Virgo-KAGRA GWTC-3 catalog, successfully inferring the population of stellar-mass binary black holes using an injection-based selection effect correction. Our results confirm that the method is robust, scalable, and applicable across different detector sensitivities and source populations.

Compressing resource-intensive large language models by removing whole transformer blocks is a seemingly simple idea, but identifying which blocks to remove constitutes an exponentially difficult combinatorial problem. In this paper, we formulate block removal as a constrained binary optimization problem that can be mapped to a physical system (Ising model), whose energies are a strong proxy for downstream model performance. This formulation enables an efficient ranking of a large number of candidate block-removal configurations and yields many high-quality, non-trivial solutions beyond consecutive regions. We demonstrate that our approach outperforms state-of-the-art block-removal methods across several benchmarks, with performance gains persisting after short retraining, and reaching improvements of up to 6 points on the MMLU benchmark. Our method requires only forward and backward passes for a few active parameters, together with an (at least approximate) Ising solver, and can be readily applied to any architecture. We illustrate this generality on the recent NVIDIA-Nemotron-3-Nano-30B-A3B-FP8 model, which exhibits a highly inhomogeneous and challenging block structure.

Genome sequencing is essential to decode genetic information, identify organisms, understand diseases and advance personalized medicine. A critical step in any genome sequencing technique is genome assembly. However, de novo genome assembly, which involves constructing an entire genome sequence from scratch without a reference genome, presents significant challenges due to its high computational complexity, affecting both time and accuracy. In this study, we propose a hybrid approach utilizing a quantum computing-based optimization algorithm integrated with classical pre-processing to expedite the genome assembly process. Specifically, we present a method to solve the Hamiltonian and Eulerian paths within the genome assembly graph using gate-based quantum computing through a Higher-Order Binary Optimization (HOBO) formulation with the Variational Quantum Eigensolver algorithm (VQE), in addition to a novel bitstring recovery mechanism to improve optimizer traversal of the solution space. A comparative analysis with classical optimization techniques was performed to assess the effectiveness of our quantum-based approach in genome assembly. The results indicate that, as quantum hardware continues to evolve and noise levels diminish, our formulation holds a significant potential to accelerate genome sequencing by offering faster and more accurate solutions to the complex challenges in genomic research.

Nekrasov's gauge origami theory provides a (complex) 4-dimensional generalization of the ADHM quiver and its moduli spaces of representations. We describe the origami moduli space as the zero locus of an isotropic section of a quadratic vector bundle on a smooth space. This allows us to give an algebro-geometric definition of the origami partition function in terms of Oh--Thomas virtual cycles. The key input is the computation of a sign associated to each torus fixed point of the moduli space. Furthermore, we establish an integrality result and dimensional reduction formulae, and discuss an application to non-perturbative Dyson--Schwinger equations following Nekrasov's work. Finally, we conjecture a description of the origami moduli space in terms of certain 2-dimensional framed sheaves on $\mathbb{P}^1 \times \mathbb{P}^1 \times \mathbb{P}^1 \times \mathbb{P}^1$, which we verify at the level of torus fixed points.

This thesis investigates the interactions of partially massless (PM) fields in 4-dimensional (anti)de Sitter spaces, along with conformal higher spin fields and their coupling to matter in arbitrary dimensions. The first part of the thesis deals with PM fields and PM algebras. A reformulation of PM fields is proposed and studied using a novel chiral formulation, inspired by Penrose's twistor approach to massless fields in Minkowski space. This reformulation enables explicit construction of Yang-Mills-type interactions and current couplings. Next, an oscillator realisation for PM higher spin algebras is given in terms of bosonic and fermionic oscillators. The construction is based on the Weyl-Clifford algebra. The second part of the thesis derives the coupling between a massless scalar field and a background of higher spin fields within a manifestly covariant framework, employing Fedosov quantization techniques, called the "parent formulation". This formalism yields, in particular, an explicit covariant expression for the coupling between scalar fields and higher spin conformal gravity.

The evolution of the vacuum state in a time-dependent external electric field of arbitrary polarization is investigated within a nonperturbative framework of quantum kinetic equations (QKEs). In our previous work [Phys. Rev. Res. 6, 043009 (2024)], a revised version of the QKEs was derived by using an adiabatic basis constructed from one-particle Hamiltonian eigenfunctions in a spatially homogeneous electric field. In this study, we present an extensive analysis of these equations with particular emphasis on observable quantities. Specifically, we compute momentum-resolved particle yields, the induced electron-positron current, the energy-momentum tensor, and the angular-momentum tensor. We also discuss in detail the charge-renormalization procedure required to remove logarithmic divergences. It is shown that our results are consistent with the previous findings obtained via the Dirac-Heisenberg-Wigner formalism. Our analysis provides a firmer theoretical basis for investigations of nonperturbative effects in strong electric fields.

In this note I review the derivation of the BK equation for the dipole gluon density in momentum space, starting from the BK equation in position space. The note contains no new results; its purpose is to consolidate derivations and formulas scattered across the literature and to establish a consistent notation.

We explore the potential for extracting proton structure functions, proton parton density functions (PDFs), and the strong coupling $α_s(M_z^2)$, using early science data from the future Electron-Ion Collider (EIC), both standalone, and in combination with HERA data. Different scenarios are considered in which samples with modest luminosity are collected at either two or three EIC beam energy configurations. The Rosenbluth separation method is used to extract the proton structure functions $F_2$ and $F_L$ from simulated data in a model-independent manner, showing that $F_L$ can be extracted significantly more precisely with three centre of mass energies than with two, whilst also obtaining $F_2$ to higher precision than has been achieved previously. The inclusion of a third beam configuration is also beneficial in the extraction of the strong coupling $α_s(M_z^2)$ that is obtainable with unprecedented experimental precision with the early EIC data. Additionally, the precision of the proton PDFs is improved when adding these data, especially for large values of Bjorken-$x$, for both two and three EIC beam energy configurations. These studies show that EIC data will already be a highly competitive probe of perturbative Quantum Chromodynamics within the first five years of data taking.

One fundamental physics interpretation of Dark Energy Spectroscopic Instrument (DESI) results is that the observed accelerated expansion of the Universe is driven not by the cosmological constant, but by a slowly rolling scalar field, a natural model for which is an axion with a decay constant close to the Planck scale. In the ``string axiverse'' one expects not one, but many, axions and this may allow the ability to engineer large effective decay constants. We thus investigate dark energy dynamics in a toy model with two axions, and compute the prior probability for the dark energy equation of state parameters $(w_0,w_a)$ implied by priors on the fundamental parameters of the model conditioned weakly on the resulting cosmology. We find various interesting behaviours, for example where both fields can be in an oscillating regime while maintaining $w_0<-1/3$, and models with opposite sign for $w_a$ than in a single-field model. The prior probability in $(w_0,w_a)$ gains support away from the ``thawing quintessence'' behaviour preferred by DESI, and when the axions have cross-interactions this effect becomes stronger. When the axions interact, one in general still requires large decay constants, but if the axions are non-interacting the largest decay constant can be reduced. The novel equations of state we illustrate may also be of relevance to models of early dark energy or inflation/reheating with mutliple axions. Further exploration of dynamics in models with $N\gg 1$ axions is warranted.

Different bases for the spin-1 density matrix are discussed to clarify the connection between its components and observables measured in heavy-ion collisions. The theoretical advantage of using the adjoint representation for spin matrices is emphasized. Next, the equilibrium spin density matrix and the corresponding Wigner function are introduced. With appropriate definitions of the energy-momentum and spin tensors, this framework allows for the formulation of perfect spin hydrodynamics in the same way as previously done for spin-1/2 particles. Together, these results provide a unified description of spin-1/2 and spin-1 particles.

A comparative description is made of the measurements at LHC of pp elastic scattering at 13 TeV by the ATLAS and TOTEM Collaborations. In the total and differential cross sections we show that the differences are justified through single numerical factor. It seems that there is no fundamental physical difference, but only a difference of normalization between the two experiments. We study the real and imaginary amplitudes disentangled with the KFK (Kohara-Ferreira-Kodama) model and show that the properties are similar in qualitative aspects for both experiments. The real and imaginary parts have different slopes at the origin and present zeros, with distributions that are common to several models, with three zeros in the real part and one zero in the imaginary amplitude. A zero in the real part, known as Martin's zero, influences the determination of the $ρ$ parameter.

We investigate the inclusive weak decay $Ξ_{bcq} \to Ξ_{ccq} + X$ ($q=u,d,s$) within the framework of the non-relativistic potential quark model. Treating the heavy diquarks $Φ_{bc}$ and $Φ_{cc}$ as compact color anti-triplet bound states, we calculate the transition form factors via the overlap integral of Schrödinger wave functions derived from the Cornell potential. The decay rates for four dominant channels-$Φ_{bc} \to Φ_{cc} + \bar{c}s$, $Φ_{bc} \to Φ_{cc} + \bar{u}s$, $Φ_{bc} \to Φ_{cc} + e/μ+ \barν$, and $Φ_{bc} \to Φ_{cc} + τ+ \barν_τ$-are computed, yielding a total inclusive decay width of $Γ(Ξ_{bcq} \to Ξ_{ccq} + X) \approx 4.1 \times 10^{-13}$ GeV. We also evaluate the potential background from $B_c^- \to Ξ_{cc} + X$ decays, finding a rate of approximately $ 3.0 \times 10^{-14}$ GeV, which is roughly one order of magnitude smaller than the signal process. Our results suggest that the inclusive decay $Ξ_{bc} \to Ξ_{cc} + X$ , followed by the well-established $Ξ_{cc}$ decay chains, provides a viable discovery channel for the doubly heavy baryon $Ξ_{bc}$ at the LHC.

These are lecture notes for an introductory course on Nichols algebras. As a main reference, I work with the book by Heckenberger and Schneider, but I want to take a distinct categorical perspective and try to develop the topic for an audience without a background in Hopf algebras. On the other hand I put some emphasis on hands-on examples. My first goal is to explain the definitions and the striking properties of Nichols algebras, foremost the odd reflection theory that is already present in Lie superalgebras. My second goal is to explain how the category of representations of a quantum group can be constructed, using categorical tools, from the Nichols algebra as its centerpiece. This makes the zoo of different existing versions of quantum groups more transparent and allows the construction of many more non-semisimple modular tensor categories. Other topics include different types of examples beyond the diagonal case, categorical versions of some Hopf algebra constructions, and an outlook section on the appearance of Nichols algebras in conformal field theory.

We propose a new class of sigma models based on Courant sigma models. We refer to these models as gauged Courant sigma models (GCSMs). By introducing additional gauge symmetries, such as those associated with a Lie group, a Lie groupoid (or Lie algebroid), and a Courant algebroid on the target space, Courant sigma models are extended to gauged sigma models of AKSZ type. The consistency of the theory is ensured by identities among geometric quantities on Lie algebroids and Courant algebroids, such as curvatures and torsions, which can be interpreted as flatness conditions on the target space. We also analyze geometric structures of GCSMs in the presence of fluxes and boundaries.

We present a comprehensive model study of the eigen-microstate approach (EMA) for identifying critical fluctuations in relativistic heavy-ion collisions. Using UrQMD and two stochastic baseline models, we demonstrate that EMA is insensitive to conventional short-range correlations and effectively filters out non-critical backgrounds. Critical fluctuations embedded via event-level or particle-level replacement with CMC events generate characteristic cluster-like eigen-microstate patterns and enhanced leading eigenvalues, with event-level criticality producing stronger responses. The eigen microstates exhibit the same pattern across different scales, demonstrating that the fractal nature of critical fluctuations is captured by the eigen microstates. Finite-size scaling of eigenvalue ratios exhibits fixed-point behavior, confirming the largest eigenvalue as an effective order-parameter-like quantity. These results demonstrate that EMA offers a robust and background-independent method for critical-point searches in the RHIC Beam Energy Scan and future heavy-ion experiments.

Although heavy-quark symmetry predicts a $B_1\bar{B}$ molecular partner of the $D_1\bar{D}$ molecule, no such state has been observed. We propose that the experimentally observed $Υ(11020)$ may be a candidate for such a state, possibly containing a $B_1\bar{B}^{*}$ component. To test this, we interpret $Υ(11020)$ as an $S$-wave $B_1\bar{B}$--$B_1\bar{B}^{*}$ molecule and compute its strong decay widths using the compositeness condition and effective Lagrangians. The couplings to $B_1$ and $\bar{B}^{(*)}$ are extracted by fitting $Υ(11020)\to e^+ e^-$ and $Υ(11020)\to χ_{bJ} πππ$ data. Using these couplings, we evaluate partial widths into $B^{(*)}_{(s)}\bar{B}^{(*)}_{(s)}$, $ππΥ(nS)$, $ππh_b(nP)$, and $πππχ_{b1}$ via hadronic loops, as well as three-body $B^{*}π\bar{B}^{(*)}$ decays via tree diagrams. The results indicate that $Υ(11020)$ is predominantly a $B_1\bar{B}$ molecule, with its main decay channel being $B_s^{*}\bar{B}^{*}$. The $ππΥ(nS)$ and $ππh_b(nP)$ widths are only a few eV, whereas $πππχ_{b1}$ reaches 0.167~MeV and the unobserved $πππχ_{b0}$ could be 0.754~keV. These distinctive decay patterns provide clear experimental signatures of the molecular nature of $Υ(11020)$ and offer a test of heavy-quark symmetry.

These proceedings present the outcome of a feasibility study using RapidSim simulation software that demonstrates that the LHCb experiment will be capable of observing the decays $B^+\to τ^+ ν_τ$ and $B_c^+\to τ^+ ν_τ$ using the data that is being collecting during Run 3 of the LHC. The proposed analysis exploits the small distance of only 5.1 millimetres between the sensing elements of LHCb's innermost silicon pixel detector, the VELO, and the LHC's proton beams to identify direct pixel hits in the VELO that can be associated with the charged $B^+$, $B_c^+$ or $τ^+$ particles. By using this extra information, the limitations due to the missing momentum and vertex information will be significantly reduced. This provides enough statistical power to pursue the measurements of these two decay channels at the LHC. In particular for the decay $B_c^+\to τ^+ ν_τ$, which has been identified by the high energy physics community as a key objective for experiments at the planned next-generation particle accelerators, this means we do not need to wait for the 2030s or beyond to get first experimental constraints.

A search is performed for a heavy scalar resonance X decaying to a Higgs boson (H) and a Higgs-like scalar boson (Y) in the two bottom quark (H $\to$ $\mathrm{b\bar{b}}$) and four quark (Y $\to$ VV $\to$ 4q) final state, where V denotes a W or Z boson. Masses of the X between 900 and 4000 GeV and the Y between 60 and 2800 GeV are considered. The search is performed in data collected by the CMS experiment at the CERN LHC from proton-proton collisions at 13 TeV center-of-mass energy, with a data set corresponding to a total integrated luminosity of 138 fb$^{-1}$. It targets the Lorentz-boosted regime, in which the products of the H $\to$ $\mathrm{b\bar{b}}$ decay can be reconstructed as a single large-area jet, and those from the Y $\to$ VV $\to$ 4q decay as either one Y $\to$ 4q or two V to $\mathrm{q\bar{q}}$ jets. Jet identification and mass reconstruction exploit machine-learning tools, including a novel attention-based "particle transformer" for Y $\to$ 4q identification. No significant excess is observed in the data above the standard model background expectation. Upper limits on the product of production cross section and branching fraction as low as 0.2 fb are derived at 95% confidence level for various mass hypotheses. This is the first search at the LHC for scalar resonances in the all-hadronic $\mathrm{b\bar{b}}$VV decay channel.

Interference effects are an important consequence of a correct description in physics theories within and beyond the Standard Model (SM) of particle physics. However, many current theoretical descriptions as well as experimental searches neglect such effects, which can, among others, lead to an incorrect description of e.g. kinematical distributions, at least within the context of UV-complete models. In this review, I briefly discuss the current status and most common descriptions as well as existing studies of such effects, where I focus on models with extended scalar searches.

Mass formulas evaluate the total weighted count of a given class of algebraic structures, such as lattices or codes. We show that 3d TQFTs provide a generalization of this concept: the total weighted count of topological boundary conditions is given by the TQFT partition function averaged over all closed 3d manifolds. This weighted count, which we call the mass, can be interpreted as the renormalized partition function of TQFT gravity. For Abelian TQFTs, the mass formula for topological boundary conditions reduces to the mass formula for particular families of codes. Focusing on the Abelian case, we show how to evaluate the mass for any bosonic theory and consider many explicit examples. We then discuss the non-Abelian generalization and compute the mass for $n + \bar n$ copies of the Ising modular tensor category. Finally, we generalize the construction to five dimensions and compute the mass for Abelian 2-form Chern-Simons theories.

We study the holographic realization of the $SO(6)_R$ R-symmetry of $\mathcal{N}=4$ super Yang-Mills with unitary gauge group. Focusing on 1/2 BPS states in the $[0,J,0]$ representation of $SO(6)_R$, it is known that depending on the scaling of $J$ with $N$, these are best described holographically in terms of gravitational waves ($J\ll N$) or D3 brane giant gravitons ($J\sim N$). These two descriptions are bridged by the dielectric effect, as the D3 giant can be regarded as a puffed-up configuration of gravitational waves. The natural non-BPS branes for symmetry operators are either 4-branes or non-BPS Kaluza-Klein monopoles. We show that the former can be regarded as a dielectric expansion of the latter, in parallel to the charged operators. We also propose symTh and symTFT candidates for the $SO(6)_R$ symmetry, whose operators at the boundary must correspond to the non-BPS branes.

In recent work we began a study of the correlators of multiple light-like Wilson loops in $\mathcal{N}=4$ super Yang-Mills theory, focussing primarily on tree-level calculations and, beyond tree-level, to the Abelian theory. Here we calculate $O(g^2)$ correlators of multiple light-like Wilson loops in the $SU(N)$ theory. We use the chiral box expansion and a study of the leading singularities of the loop integrand to arrive at integrated expressions for these objects. We then use the results of these calculations to verify that a natural generalisation of the $\bar{Q}$-equation, familiar from the study of single Wilson loops, holds in the $SU(N)$ theory. This $\bar{Q}$-equation should provide a valuable tool for the computation of multiple Wilson loop correlators at higher order in the coupling.

We discuss the Kontsevich-Segal-Witten criterion for the allowability of complex metrics, in the context of the gravitational path integral that calculates the supersymmetric index. We focus on the saddle points that capture the contribution of supersymmetric black holes in AdS$_5$ space. We show that, for such black holes with two independent angular momenta, the conditions imposed on the corresponding saddle point by the KSW criterion are equivalent to the ones arising from the convergence of the microscopic trace form of the supersymmetric index. This result adds to previous results establishing such an equivalence in other, simpler examples of the gravitational index in AdS space and flat space. Along the way, we give a practical algorithm for implementing the KSW criterion in terms of eigenvalues of certain matrices.

We show how various constructions of $\mathbb{Z}$-graded Lie superalgebras are related to each other. These Lie superalgebras have a Lie algebra $\mathfrak{g}$ as the subalgebra at degree 0, an odd $\mathfrak{g}$-module V as the subspace at degree 1, and an embedding tensor as an element at degree -1. This is a linear map from V to $\mathfrak{g}$ satisfying a quadratic constraint, which equips V with the structure of a Leibniz algebra.

Stochastic hydrodynamics provides a dynamical framework for the evolution of fluctuations in heavy-ion collisions, but poses significant challenges in numerical simulations. We present an algorithm for the simulation of non-relativistic stochastic fluids in two spatial dimensions in a box. We use the robust Metropolis algorithm, handling fluctuations and dissipation at once by systematically replacing dissipative terms in the hydrodynamic equations by random forces. The algorithm can easily be modified for numerical simulations of other hydrodynamic theories. We present test cases as well as numerical calculations of the renormalization of shear viscosity.

This paper explores the catastrophic energy transformations, in particular the ones leading to the generation of a flow in a weakly rotating self-gravitating fluid/gas found, for instance, in the vicinity of a massive compact object. Because of the similarity in the governing equations, the system dynamics is worked out exactly in parallel to the methods developed for investigating catastrophic relaxation in stellar plasmas [1-3]. In the latter a more ``complex" equilibrium state, on slow changes in the environment, can lose its equilibrium (catastrophe), and transform to a less complex state with a very different energy mix from the original. It is shown that a similar transformation in the weakly rotating self-gravitating fluid/gas will convert much of its gravitation energy into kinetic energy in the flow. Since flows are a perennial ingredient of high-energy astrophysical systems, the energy transformation processes revealed in present study, can advance our understanding of a variety of them. Some particularly relevant examples are: macro-scale flows / structures in galaxies, accretion discs, and the dynamics and stability of a rotating star / its atmosphere.

Spectroscopic surveys allow spatially resolved spectroscopy of galaxies to study their interstellar medium (ISM). However, observations of Galactic H$α$ emission are contaminated by geocoronal H$α$ emission. The latter is known to depend on the shadow altitude, a geometric parameter relating the line of sight to Earth's shadow cone. Using fibres on blank skys from the SDSS-IV/MaStar survey, we established an empirical relation between the geocoronal H$α$ emission and the shadow altitude, with a root mean square fractional scatter of 23.52$\%$. This relation can be used to predict geocoronal H$α$ emission so that it can be removed from observed spectra. This removal method is advantageous when the observed targets are extensive in the sky, and it does not require a large velocity separation between the observed target and the local standard of rest. This will enable reliable studies of Galactic H$α$ in intermediate spectral resolution integral field spectroscopic surveys. We also find tentative evidences for the dependences of geocoronal emission on solar activity and the distance between the Earth and the Sun.

A new pulsar signal-processing pipeline has been developed for observing pulsars with the Diamond Array Radio Telescope at the Gauribidanur radio observatory. The array consists of 32 off-axis dual-polarised LPDAs, with a nominal gain of 22 dBi between 130 and 350 MHz and a 15-degree HPBW at 175 MHz for transit observations on pulsars. Custom-built receivers and real-time data-capture and analysis tools have been developed and used. Receiver output voltages from a transient buffer, as well as full-polar spectral data at both high and low resolutions, suitable for transient searches and pulsar studies. Additionally, full-polar folded profile archives are generated for known pulsars in subintegrations and both coherent and incoherent dedispersion. Custom-developed Python routines, FFT libraries, DSPSR, PSRCHIVE, and Presto modules have been used to build the pipeline. The functionalities of the pipeline were validated with artificially generated pulsar signals and strong celestial sources before it was released for routine observations. Presently, the pipeline is configured to observe pulsars between 170 and 196 MHz, with a daily cadence. Recorded data are reduced in-line immediately following each observation, nearly matching the observation time at a 1:1 ratio. An Intel i9 server captures the data, and an AMD R9 CPU does the primary data reduction. The archives are routinely backed up to a remote system via the internet. The paper presents the architecture of the signal processing pipeline developed, its validation, and initial polarimetric results observing five bright pulsars at 175 MHz. Results also include RM estimates and single-pulse study results for B0953+08, B0531+21, and B1133+16, as well as from monitoring the spin-down of the Crab pulsar over 200 days of observation. Finally, it presents a discussion on the potential improvements for the array.

Double white dwarf (DWD) binaries are natural outcomes of binary stellar evolution and key sources for future space-based gravitational wave (GW) observatories such as Laser Interferometer Space Antenna (LISA). We investigate how different binary interaction channels shape the physical and orbital properties of DWD systems, focusing on component masses, orbital separations, core compositions, and mass transfer rates. Using the binary population synthesis code COMPAS, we evolve $10^7$ binaries with physically motivated initial distributions of binary parameters. Our simulations reproduce the strong bimodality in the final orbital separations, including a pronounced deficit of systems around $100-500 \rm\,R_\odot$, arising from distinct evolutionary pathways: wide DWDs predominantly originate from stable Roche lobe overflow (RLOF), while close DWDs form through unstable RLOF leading to at least one common envelope (CE) phase. Moreover, we show that the core compositions of WDs provide a powerful tracer of evolutionary history: He-core WDs are strongly concentrated in close systems, whereas CO-core WDs span the full separation range and exhibit a small mass gap in wide binaries. We further identify a correlation between the accreted mass and the final orbital separation, highlighting the impact of non-conservative mass transfer on the resulting orbital configuration of DWD systems. These results underscore the links among evolutionary channels, chemical composition, and mass transfer rates; thereby provide a unique framework for interpreting Gaia DWD samples and forecasting the joint electromagnetic and GW population accessible to LISA.

We consider the Single-Mirror Small-Size imaging atmospheric Cherenkov Telescopes (SST-1M) to be located inside a high-altitude array of Water-Cherenkov Detectors (WCDs) inspired by the Southern Wide-field Gamma-ray Observatory (SWGO). For such a hybrid observatory, using detailed Monte Carlo simulations, we show an improvement in the flux sensitivity of monocular and stereoscopic SST-1M observation by about 60% and 30% above 10 TeV, respectively, due to the improved gamma/hadron separation when additional parameters from the WCD array are used. We also discuss further benefits of the hybrid SWGO concept and its technical challenges.

Gauribidanur Diamond Array Radio Telescope (GBD-DART) is a new small LPDA antenna array consisting of 64 short dipoles and associated receivers that has been custom developed and deployed at the Gauribidanur observatory (13.604 N, 77.427 E) to study bright Pulsars and Solar transients in the frequency range of 130-350 MHz. The LPDAs are arranged in a checkerboard layout, with opposite pairs combined to enable dual-polarised operation. A diamond-shaped (tilted square) array configuration was chosen to achieve high sidelobe suppression in the East-West and North-South directions. The tile measures 5.9 meters by 5.9 meters, with diagonals along both the North-South and East-West directions, each measuring about 8.4 meters. The LPDA array with one diamond-shaped tile has been fully commissioned and is operating in transit-observing mode, successfully detecting strong pulsars and solar flares over the last seven months. The present digital backend restricts the instantaneous bandwidth for observations to 16 MHz. The array operations are streamlined to facilitate remote operations. Apart from investigating Pulsar and Solar phenomena at low radio frequencies in selected sources, this work aims to provide a training platform for radio astronomy through simple-to-construct, low-cost radio telescopes. In this paper, we present details of the array, including antenna and array response studies, brief descriptions of front-end and backend instrumentation, and illustrative results from observations of both pulsars and solar flares. It will also provide brief details of future upgrade plans, particularly for the tiles and digital backend, to facilitate the observation of additional sources.

A significant fraction of gamma-ray bursts (GRBs) exhibit a plateau in the early X-ray afterglow light curve, whose mechanism remains uncertain. While the post-plateau normal decay index ($α_2$) is commonly used to constrain the afterglow dynamics, the shallow-decay slope of the plateau itself ($α_1$) has received comparatively little attention. Recent observations, however, reveal substantial dispersion in $α_1$, raising the question of whether GRBs with rising, flat and mildly decaying plateaus represent intrinsically distinct populations. To address this question, we collect a uniform sample of 185 $\textit{Swift}$ GRBs with a well-defined plateau and divide them into three groups based on $α_1$. Using a non-parametric approach, we reconstruct their X-ray luminosity functions, redshift distributions and event rates. It is found that the three groups exhibit statistically consistent properties across all diagnostics, with no evidence for group-specific features. Monte Carlo perturbation tests further show that these results are insensitive to the adopted classification boundaries of $α_1$. Our results indicate that variations in the plateau slope $α_1$ do not define distinct GRB subclasses, but instead the sample constitutes a statistically uniform population governed by a common framework.

The exponential profile has long been hypothesized as the fundamental morphology of galactic disks. The IllustrisTNG simulations reproduce diverse surface-density profiles: Type I (single exponential), Type II (down-bending), and Type III (up-bending), consistent with observed mass-size relations and kinematics. Type II disks dominate the stellar-mass regime $M_\star < 10^{10.6} M_\odot$ with a prevalence of about 40%, exhibiting systematically extended morphologies. Conversely, Type III and Type I galaxies are more compact while following the same mass-size scaling relation. Evolutionary histories show that Type II galaxies experience minimal external perturbations, suggesting that Type II disks represent an intrinsic disk form and challenging conventional single-exponential paradigms. We demonstrate that Type II breaks arise naturally via inside-out growth since $z=1$, governed by synchronized cold-gas accretion and localized peaks in specific star formation rate. This mechanism also produces the characteristic U-shaped age profiles of Type II disks. Stellar dynamical redistribution plays a minor role in their formation.

Star clusters are key components of galaxies, and the relationship between cluster radius and mass encodes information about cluster formation and evolution. Theoretical models predict that age and specific star formation rate (sSFR) should influence cluster size through stellar mass loss and gas dynamics during formation. We hypothesized that if these theoretical predictions hold, multivariate models including age and sSFR should predict cluster radius better than models using mass alone. To test this, we used regression analysis on 5,105 star clusters from the LEGUS survey, comparing a full multivariate model against a mass-only baseline. We found that mass dominated the radius-mass relation: log(Mass) showed a strong correlation with radius (coefficient = 0.131 +/- 0.008, p < 0.001), while log(sSFR) and log(Age) contributed negligibly (0.0002 +/- 0.015 and 0.038 +/- 0.006, respectively). Cross-validation revealed that the mass-only model generalized better (CV R^2 = 0.028 vs -0.017), with the negative value for the multivariate model indicating overfitting. Contrary to our hypothesis, adding age and sSFR did not improve predictive performance. The low R^2 (0.115) indicated that most variance in cluster radius remained unexplained by these variables, suggesting other factors may play important roles. Among the variables tested, our findings were consistent with virial equilibrium predictions, with mass serving as a more fundamental parameter than evolutionary age or galaxy star formation rate.

Interacting transients probe mass loss in the final stages of stellar evolution; however, the geometry and timing of multi-episode mass loss remain poorly constrained. SN 2024abvb is a nearby interacting event with transitional Ibn/Icn spectroscopic properties and multi-epoch polarimetry, offering a rare opportunity to study structured circumstellar material (CSM). We aim to characterise the kinematics, composition and geometry of the CSM around SN 2024abvb and to identify plausible progenitor/ejection scenarios that can produce the observed spectro-polarimetric evolution. We present high-resolution (VLT/UVES and VLT/X-Shooter) optical/NIR spectroscopy across several epochs, complemented by broadband polarimetry and spectropolarimetry (VLT/FORS2 and NOT/ALFOSC). Line identifications, velocity decompositions and polarimetric time-series are used to trace multiple kinematic components and changes in scattering geometry. The high-resolution spectra reveal multiple narrow CSM components composed of He, C and O with absorption minima at $\sim150 - 400$ km s$^{-1}$ and additional faster material up to $\sim2000$ \kms. Low-velocity Balmer absorptions are present, indicating distant H-rich material, a first in SNe Ibn/Icn. Polarimetry shows a marked evolution ($P\sim1\%$ near peak, $\lesssim0.5\%$ after $\sim1$ week, rising to $\sim1.5\%$ at $\sim20$ d with $\sim50^\circ$ position-angle rotation and to $\sim4\%$ at $\sim30$ d, stronger in the blue), implying a time-variable, wavelength-dependent scattering/obscuration component. The combination of kinematics and polarimetric behaviour is consistent with multiple, concentric toroidal shells with differing orientations and partial dust content.

We present the results from a five-year campaign to gather multi-epoch, high-resolution spectra of a volume-complete sample of 413 M dwarfs with masses 10-30% that of the Sun that lie within 15 parsecs. We report weighted mean systemic radial velocities (RV) and rotational broadening measurements ($v \sin i$) for our targets. Our typical relative RV uncertainties are less than $50$ m/s for the isolated, slowly rotating targets in our sample, and increase but remain less than 1 km/s for more rapidly rotating stars. The majority of the single stars in our sample ($71\pm3$%) have rotational broadening below our detection limit of 3.4 km/s. When combined with astrometric data, our radial velocities allow us to calculate galactic space motions, which we use to assign thin or thick disk membership. We determine that 81% and 8% of our sample are highly probable thin and highly probable thick disk members, respectively. We report seven new multi-lined multiple systems and identify six additional targets with velocity variations indicative of long-period binaries, of which three are new detections. Finally, we find no significant difference in the stellar multiplicity rates of the thin disk ($22\pm2$%) and thick disk ($21\pm8$%) populations in our sample, implying that mid-M dwarfs are not significantly losing their companions at these relative ages. Our survey more than triples the number of these fully-convective stars with complete astrometric data and uniformly derived, multi-epoch, high-resolution RVs and rotational broadening measurements.

Recent works show that the radio power of quasars accreting at very high rates can reach surprisingly high values. These studies suggest that this radio emission might originate from star formation, but lack of data leaves open the possibility that they could also contain a jetted active galactic nucleus (AGN). We investigate the origin of the radio emission of a sample of 18 super-Eddington candidates, over a wide range of redshifts. These sources are expected to have extreme radiative output per unit black hole mass, show high-velocity outflows and are therefore thought to be a prime mover of galactic evolution via radiative and mechanical feedback. We present new Karl G. Jansky Very Large Array (VLA) observations at L, C and X-band of these sources, which we combine with observations from the LOw-Frequency ARray (LOFAR) Two-metre Sky Survey (LoTSS) and the Very Large Array Sky Survey (VLASS). We also use optical and IR data to derive estimates of accretion and wind parameters, as well as star formation rates to compare with the ones derived from the radio emission. Based on the radio variability, luminosity, morphology, radio spectral properties, radio vs IR estimates of star formation rate and radio-to-mid IR flux ratio, we find that 7 of our 18 targets are likely to have their radio emission predominantly coming from SF, and 6 from a combination of SF and AGN-related mechanisms, while only three sources indicate a core or jetted AGN only origin for the detected radio emission. This is consistent with previous studies, and supports the prevalence of lower power radio structures associated with star-forming activity rather than relativistic jets in the high Eddington ratio regime. In the same sample, however, we find three sources for which the data suggest a concomitant presence of super-Eddington accretion and relativistic ejections.

The avalanche mechanism of plasma production in active galactic nuclei (AGNs) is detailed, and constraints on system parameters needed for efficient electron-positron pair cascades are explored. Whether an AGN falls within this favorable parameter range may explain the observed radio-loud versus radio-quiet dichotomy. On the other hand, this study shows that cascades generate orders of magnitude fewer pairs than is necessary to explain the synchrotron emission observed in luminous jets. This fact suggests the existence of either an alternative lepton source, namely pair production of photons from the hot accretion flows around AGN central black holes, or matter loading of the jets from the surrounding medium, or, most likely, both. The case of the radio galaxy 3C 120 is considered in detail.

The ultra-faint dwarf galaxies (UFDs) are some of the oldest and most metal-poor environments in the Local Group. In particular, they are predicted to host the first stars (only H and He) that lit up in our Universe. No metal-free stars have been found to date, but their chemical products can be observed on the surfaces of the ancient second-generation stars such as the carbon-enhanced metal-poor stars (CEMP-no, [C/Fe]>+0.7).} However, in each UFD there are only a few stars bright enough for spectroscopic follow-up, therefore it is crucial to study as many of these systems as possible. Here we follow up stars belonging to two recently discovered UFDs, Grus II and Tucana IV. The spectra analyzed were obtained with the multi-object spectrograph FLAMES/Giraffe at the Very Large Telescope (VLT). This includes spectra in two wavelength ranges: red spectra around the CaII triplet (8498 Å, 8542 Å, 8662 Å) used to derive radial velocity and [Fe/H], and blue spectra covering the CH band at ~ 4300 Å. In total, we analyzed 21 spectra of member candidates for Grus II and 17 for Tucana IV, including both Red Giant Branch (RGB) and Horizontal Branch (HB) stars. We identified 13 members in Grus II (thereof 8 RGB stars) and 7 members in Tucana IV (thereof 3 RGB stars). Among the RGB stars in Grus II, we found three CEMP-no stars at [Fe/H]~-3 and [C/Fe]>+1 and two CEMP-no stars at slightly higher [Fe/H] and [C/Fe]>+0.7. In Tucana IV, we found one CEMP-no star ([Fe/H]=-2.75 and [C/Fe] = +0.83). This project, along with future investigations of CEMP stars in UFDs, allows us to study the impact of the first stars in these ancient and primitive systems and consequently the first chemical enrichment that occurred in the Universe.

The physical conditions of the intergalactic medium (IGM) during the final stages of cosmic reionization ($z\sim5.0-6.0$) are not yet fully understood. Recent reports of unexpectedly large-scale ($\ge 150 h^{-1}\mathrm{cMpc}$) correlation in Ly-$α$ transmission flux using extended XQR-30 quasar spectra pose interesting consequences on the reionization end stages. In this work, we investigate the Ly-$α$ forest dark-gap distribution (defined as regions with transmitted flux below 0.05) as another sensitive tracer of the IGM, using an efficient, large-volume ($\sim 1 ~\mathrm{Gpc}$) simulation framework. By constructing a suite of physically motivated model variants (i.e, varying the reionization redshift, IGM temperature, and ionizing-photon mean free path), we generate synthetic sightlines and compare their predicted cumulative distribution of dark gaps with that of observed spectra (at redshift intervals of $Δz=0.2$). We find that most of the models achieve qualitatively consistent agreement with the data. Specifically, the scenario involving a slightly later reionization completion ($z\sim 5.4$) provides the closest match, while a short constant mean free path model disfavors the data at lower redshifts. These findings give further support for the emerging scenario of reionization end extending to $z\le5.7$, although they can not rule out a slightly early reionization with enhanced post-ionization ultraviolet (UV) background fluctuations. A similar conclusion arises from the redshift distribution of long dark gap ($L\ge 30 ~h^{-1}\mathrm{cMpc}$) fraction. However, the model variants are still not able to reproduce the observed strong flux correlations at unusually large scales, which remains open for further investigations.

We present an updated catalog of TeV gamma-ray sources based on the fifth pass of data from the High-Altitude Water Cherenkov (HAWC) Observatory. This release benefits from improved event reconstruction and nearly three additional years of observations. It also incorporates a systematic multi-source fitting framework, enabling more flexible and accurate modeling of the gamma-ray sky. This fitting procedure was modeled after the manual approach used in HAWC analyses of individual sources and regions, as well as other gamma-ray catalogs, like the 4FGL. In addition to more varied modeling of source morphology and spectral parameters compared to previous HAWC catalogs, this catalog uses a robust modeling of Galactic diffuse TeV emission. The fitting procedure uses both point-like and symmetric Gaussian spatial templates to model the source morphology. The spectral shape of the emission is modeled with either a simple power-law or log-parabola to explore curvature in the spectral energy distribution. We report 85 sources at the 4σ level, including 11 sources not associated with any TeVCat source using a distance-based association criterion. Distance-based association with the 1LHAASO catalog results in 22 4HWC sources without a counterpart. Additionally, there are 12 sources not associated with any physical counterpart in the Low- or High-Mass X-Ray Binary, the ATNF, or Fermi Pulsar, or SNR catalogs of sources. Five of the aforementioned sources have no counterpart in any of the catalogs searched and represent an opportunity for follow-up observations.

Quasi-periodic oscillations (QPOs) in gamma-ray bursts (GRBs) afterglows have been suggested as probes of the central engine. Such interpretations generally assume that the observed modulation frequency directly corresponds to an intrinsic oscillation frequency of the source. We show that this assumption is not generally valid and that interpreting such features without accounting for relativistic propagation may lead to misleading inferences about the engine nature. We show that relativistic propagation effects - most importantly integration over equal-arrival-time surfaces - act as a frequency-dependent filter that can significantly modify or suppress intrinsic variability. In the constant- $Γ$ case, the angular kernel acts as a stationary low-pass filter that suppresses high-frequency variability without altering its frequency, whereas Blandford-McKee deceleration renders the filter time-dependent and manifests observationally as an apparent frequency drift.

We investigate the evolution of molecular clouds through the kinematics of their atomic hydrogen (HI) envelopes, using $^{12}\mathrm{CO}$ and 21-cm emission to trace the molecular and atomic gas, respectively. We measure the large-scale gradients, $Ω$, in the velocity fields of 22 molecular clouds and their HI envelopes, then calculate their specific angular momenta, $j\propto ΩR^2$. The molecular clouds have a median velocity gradient of $9.6\times 10^{-2}\ \mathrm{km\ s^{-1}\ pc^{-1}}$, and a typical specific angular momentum of $2.7 \times 10^{24}\ \mathrm{cm^2\ s^{-1}}$. The HI envelopes have smaller velocity gradients than their respective molecular clouds, with an average of $Ω_\mathrm{HI} = 0.03\ \mathrm{km\ s^{-1}\ pc^{-1}}$, and a median angular momentum of $j_\mathrm{HI} \approx 5.7 \times 10^{24}\ \mathrm{cm^2\ s^{-1}}$. For a majority of the systems, $j_\mathrm{HI} > j_\mathrm{H_2}$, with an average of $j_\mathrm{HI}/j_\mathrm{H_2} = 4$. Their velocity gradient directions tend to be misaligned, indicating that angular momentum is not conserved during molecular cloud formation. Both populations exhibit a $j-R$ scaling consistent with that expected of supersonic turbulence: $j_\mathrm{H_2} \propto R^{1.67\pm 0.22}$, and $j_\mathrm{HI} \propto R^{1.71\pm 0.27}$. Combining our measurements with previous observations, we demonstrate a scaling of $j \propto R^{1.50\pm 0.02}$ in star-forming regions spanning 5 dex in size, $R\in (10^{-3},\ 10^2) \ \mathrm{pc}$. We construct a model of angular momentum transport during molecular cloud formation, and derive the angular momenta of the progenitors to the present-day systems. We calculate a typical angular momentum redistribution timescale of 13 Myr, comparable to the HI envelope free-fall times.

Gaia DR4 epoch astrometry will enable the detection of thousands of exoplanets through astrometric motion. Here, we present a composite catalog and modeling framework that extracts the maximum information from existing Hipparcos and Gaia data releases. We calibrate Gaia DR2 proper motions and DR3-DR2 scaled position differences against the Gaia DR3 reference frame, and combine these with the Hipparcos-Gaia Catalog of Accelerations, the Hipparcos intermediate astrometric data, Gaia astrometric excess noise, and Gaia radial velocity variability constraints. We implement a joint likelihood model for these data in Octofitter that marginalizes over Gaia's unpublished observation epochs. This results in full orbit posteriors that can be computed uniformly for a large class of companions. We compare these posteriors to published orbital solutions for 25 stellar binaries from the Sixth Catalog of Orbits of Visual Binary Stars, recovering all companions at high significance and broadly consistent orbital separations. We then recover independent evidence to support 94 of 120 tested Jovian exoplanetary systems from the NASA Exoplanet Archive (plus 3 known stellar companions, and one previously detected planet we now rule out). We demonstrate that in cases like 14 Her b, the posteriors confirm the planetary nature of a signal using only Gaia and Hipparcos data. We find no false positives among 25 RV-quiet standard stars without significant Hipparcos-Gaia accelerations. Our method can break degeneracies inherent to proper motion anomaly or excess noise modeling alone by resolving orbital curvature within the Gaia baseline. The catalog and updated Octofitter are made publicly available to the community.

We present observations of the W51A region, including the massive protoclusters W51-E and W51-IRS2, with JWST in 10 NIRCam and 5 MIRI filters. In this work, we highlight the most novel features apparent in these images and compare them with other multi-wavelength images. The broad view of the NIRCam/MIRI images of the W51A region shows that areas dominated by warm dust and ionized gas are distinct from those dominated by PAHs. The high angular resolution of the JWST images resolves dust filaments in high contrast, revealing geometrically converging features feeding W51-E and a cavity around W51-IRS2. This picture adds support to the hypothesis that feedback from W51-IRS2 is suppressing further gas infall onto the protocluster, while by contrast, gas is still accreting onto W51-E. Comparing the NIRCam and MIRI images to ALMA data, we find 24 sources detected by both JWST and ALMA, accounting for only $\sim10\%$ of the ALMA sources; the rest are too embedded or too cool to be detected by JWST. A knot of [Fe II] and H$_2$ emission north of W51-IRS2, previously detected in ground-based images, reveals peculiarly bright and compact peaks detected in all JWST bands. The knot is likely the most energetic example of a protostellar jet driven by a massive star impacting dense interstellar medium. The new images provide a complementary view to the previous long-wavelength perspective on this 4 x 8 pc area of one of the most active star-forming regions in our Galaxy, revealing new mysteries to be further explored.

The accretion of Gaia-Enceladus/Sausage (GES) onto the Milky Way (MW) is one of the most prominent features of the Galactic halo revealed by the combination of the Gaia satellite and large spectroscopic surveys. This massive accretion largely contributes to the local stellar halo mass and was significant enough to alter the formation history and the morphology of the MW. In this work, we aim to analyse the selection of stars previously identified as belonging to GES with different kinematics and chemical properties to test the hypothesis of a two-phase accretion event. We apply several statistical tests to assess the significance of the separation between the two populations in GES. We then employ galactic chemical evolution models to investigate the origin of the chemical differences encountered in the analysis. We confirm the presence of two distinct populations, with consistently different dynamical and chemical properties. The low energy population seems to show higher overall abundances, whereas the high-energy one may be more metal-poor. We attribute this difference to the presence of at least two separate populations of stars within Gaia-Enceladus, likely associated with the innermost (low-energy) and outermost (high-energy) regions of the progenitor. The adopted models successfully reproduce the patterns in metallicity and [alpha/M] distributions in an inside-out scenario. Our analysis supports the presence of a former metallicity gradient in Gaia-Enceladus, and reinforces the interpretation of its accretion as a multi-passage event through the Milky Way disc.

Observations indicate two nested pairs of extended bipolar bubbles emanating from the Milky-Way center - the $|b|\sim80^\circ$ latitude eROSITA bubbles (RBs), encompassing the smaller, $|b|\sim 50^{\circ}$ Fermi bubbles (FBs) - and classify the edges of both bubble pairs as strong forward shocks. Identifying each bubble pair as driven by a distinct, collimated outburst, we evolve these bubbles and constrain their origin using a stratified 1D model verified by a suite of 2D and 3D hydrodynamic simulations which reproduce X-ray observations. While the RBs are at the onset of slowdown, the FBs are still expanding ballistically into the RB-shocked medium. Observational constraints indicate that both RB and FB outbursts had (up to factor $\sim2$-$3$ uncertainties) $\sim4^\circ$ half-opening angles and $\sim 2000$ km s$^{-1}$ velocities $100$ pc from their base, carrying $\sim10^{55}$ erg. The FBs and RBs could thus arise from identical outbursts separated by $\sim10$ Myr; their longitudinal asymmetry favors an eastern ambient-density gradient over western wind suggestions.

We present volumetric rates and luminosity functions (LFs) of Type Ia supernovae (SNe Ia) from the All-Sky Automated Survey for Supernovae (ASAS-SN), covering the 11-year period from 2014 to 2024. By combining the 2014--2017 $V$-band sample with the 2018--2024 $g$-band sample, we construct a large statistical dataset of $1776$ SNe Ia. We compute completeness corrections based on injection-recovery simulations of the ASAS-SN light curves, taking into account the variations in light curve shapes. For our standard sample ($M_{g,\mathrm{peak}}<-16.0$ mag), we extract a total volumetric SN Ia rate of $R_{\mathrm{tot}} = (2.55 \pm 0.12) \times 10^4\,\mathrm{yr}^{-1}\,\mathrm{Gpc}^{-3}\,h_{70}^3$ at a median redshift of $z=0.029$. With a statistical uncertainty of $4.7\%$, this is the most precise local measurement to date. While the "normal" SNe Ia account for $(92.7 \pm 1.9)\%$ of this rate, the total LF reveals immense diversity, with $M_{g,\mathrm{peak}}$ spanning over five magnitudes. The LF of SNe Iax is also broad and rises toward lower luminosities, resulting in a likely lower limit of $(4.3 \pm 1.8)\%$ of the total rate. We place strong constraints on the rate of SNe Ia-CSM, finding they account for only $(0.036 \pm 0.017)\%$ of the total local rate. Finally, we find that the low-luminosity 02es-like SNe are $7 \pm 5$ times more common than the luminous 03fg-like SNe. This places demographic constraints on models proposing a physical continuum for these two subtypes, implying that any common channel for the two classes must strongly favor lower-luminosity explosions.