arXiv Daily Digest - 2026-01-14

In this work, we explore the Large Language Model (LLM) agent reviewer dynamics in an Elo-ranked review system using real-world conference paper submissions. Multiple LLM agent reviewers with different personas are engage in multi round review interactions moderated by an Area Chair. We compare a baseline setting with conditions that incorporate Elo ratings and reviewer memory. Our simulation results showcase several interesting findings, including how incorporating Elo improves Area Chair decision accuracy, as well as reviewers' adaptive review strategy that exploits our Elo system without improving review effort. Our code is available at https://github.com/hsiangwei0903/EloReview.

Despite the rapid progress of video generation models, the role of data in influencing motion is poorly understood. We present Motive (MOTIon attribution for Video gEneration), a motion-centric, gradient-based data attribution framework that scales to modern, large, high-quality video datasets and models. We use this to study which fine-tuning clips improve or degrade temporal dynamics. Motive isolates temporal dynamics from static appearance via motion-weighted loss masks, yielding efficient and scalable motion-specific influence computation. On text-to-video models, Motive identifies clips that strongly affect motion and guides data curation that improves temporal consistency and physical plausibility. With Motive-selected high-influence data, our method improves both motion smoothness and dynamic degree on VBench, achieving a 74.1% human preference win rate compared with the pretrained base model. To our knowledge, this is the first framework to attribute motion rather than visual appearance in video generative models and to use it to curate fine-tuning data.

The evolution of recommender systems has shifted preference storage from rating matrices and dense embeddings to semantic memory in the agentic era. Yet existing agents rely on isolated memory, overlooking crucial collaborative signals. Bridging this gap is hindered by the dual challenges of distilling vast graph contexts without overwhelming reasoning agents with cognitive load, and evolving the collaborative memory efficiently without incurring prohibitive computational costs. To address this, we propose MemRec, a framework that architecturally decouples reasoning from memory management to enable efficient collaborative augmentation. MemRec introduces a dedicated, cost-effective LM_Mem to manage a dynamic collaborative memory graph, serving synthesized, high-signal context to a downstream LLM_Rec. The framework operates via a practical pipeline featuring efficient retrieval and cost-effective asynchronous graph propagation that evolves memory in the background. Extensive experiments on four benchmarks demonstrate that MemRec achieves state-of-the-art performance. Furthermore, architectural analysis confirms its flexibility, establishing a new Pareto frontier that balances reasoning quality, cost, and privacy through support for diverse deployments, including local open-source models. Code:https://github.com/rutgerswiselab/memrec and Homepage: https://memrec.weixinchen.com

The Contract Net Protocol (1980) introduced coordination through contracts in multi-agent systems. Modern agent protocols standardize connectivity and interoperability; yet, none provide formal, resource governance-normative mechanisms to bound how much agents may consume or how long they may operate. We introduce Agent Contracts, a formal framework that extends the contract metaphor from task allocation to resource-bounded execution. An Agent Contract unifies input/output specifications, multi-dimensional resource constraints, temporal boundaries, and success criteria into a coherent governance mechanism with explicit lifecycle semantics. For multi-agent coordination, we establish conservation laws ensuring delegated budgets respect parent constraints, enabling hierarchical coordination through contract delegation. Empirical validation across four experiments demonstrates 90% token reduction with 525x lower variance in iterative workflows, zero conservation violations in multi-agent delegation, and measurable quality-resource tradeoffs through contract modes. Agent Contracts provide formal foundations for predictable, auditable, and resource-bounded autonomous AI deployment.

The recent development of Large Language Models (LLMs) with strong reasoning ability has driven research in various domains such as mathematics, coding, and scientific discovery. Meanwhile, 3D visual grounding, as a fundamental task in 3D understanding, still remains challenging due to the limited reasoning ability of recent 3D visual grounding models. Most of the current methods incorporate a text encoder and visual feature encoder to generate cross-modal fuse features and predict the referring object. These models often require supervised training on extensive 3D annotation data. On the other hand, recent research also focus on scaling synthetic data to train stronger 3D visual grounding LLM, however, the performance gain remains limited and non-proportional to the data collection cost. In this work, we propose a 3D visual grounding data pipeline, which is capable of automatically synthesizing 3D visual grounding data along with corresponding reasoning process. Additionally, we leverage the generated data for LLM fine-tuning and introduce Reason3DVG-8B, a strong 3D visual grounding LLM that outperforms previous LLM-based method 3D-GRAND using only 1.6% of their training data, demonstrating the effectiveness of our data and the importance of reasoning in 3D visual grounding.

Large language models often solve complex reasoning tasks more effectively with Chain-of-Thought (CoT), but at the cost of long, low-bandwidth token sequences. Humans, by contrast, often reason softly by maintaining a distribution over plausible next steps. Motivated by this, we propose Multiplex Thinking, a stochastic soft reasoning mechanism that, at each thinking step, samples K candidate tokens and aggregates their embeddings into a single continuous multiplex token. This preserves the vocabulary embedding prior and the sampling dynamics of standard discrete generation, while inducing a tractable probability distribution over multiplex rollouts. Consequently, multiplex trajectories can be directly optimized with on-policy reinforcement learning (RL). Importantly, Multiplex Thinking is self-adaptive: when the model is confident, the multiplex token is nearly discrete and behaves like standard CoT; when it is uncertain, it compactly represents multiple plausible next steps without increasing sequence length. Across challenging math reasoning benchmarks, Multiplex Thinking consistently outperforms strong discrete CoT and RL baselines from Pass@1 through Pass@1024, while producing shorter sequences. The code and checkpoints are available at https://github.com/GMLR-Penn/Multiplex-Thinking.

Tracklet quality is often treated as an afterthought in most person re-identification (ReID) methods, with the majority of research presenting architectural modifications to foundational models. Such approaches neglect an important limitation, posing challenges when deploying ReID systems in real-world, difficult scenarios. In this paper, we introduce S3-CLIP, a video super-resolution-based CLIP-ReID framework developed for the VReID-XFD challenge at WACV 2026. The proposed method integrates recent advances in super-resolution networks with task-driven super-resolution pipelines, adapting them to the video-based person re-identification setting. To the best of our knowledge, this work represents the first systematic investigation of video super-resolution as a means of enhancing tracklet quality for person ReID, particularly under challenging cross-view conditions. Experimental results demonstrate performance competitive with the baseline, achieving 37.52% mAP in aerial-to-ground and 29.16% mAP in ground-to-aerial scenarios. In the ground-to-aerial setting, S3-CLIP achieves substantial gains in ranking accuracy, improving Rank-1, Rank-5, and Rank-10 performance by 11.24%, 13.48%, and 17.98%, respectively.

We introduce the AI Productivity Index for Software Engineering (APEX-SWE), a benchmark for assessing whether frontier AI models can execute economically valuable software engineering work. Unlike existing evaluations that focus on narrow, well-defined tasks, APEX-SWE assesses two novel task types that reflect real-world software engineering work: (1) Integration tasks (n=100), which require constructing end-to-end systems across heterogeneous cloud primitives, business applications, and infrastructure-as-code services, and (2) Observability tasks (n=100), which require debugging production failures using telemetry signals such as logs and dashboards, as well as unstructured context. We evaluated eight frontier models on APEX-SWE. Gemini 3 Pro (Thinking = High) performs best, with a Pass@1 score of 25\%. Our analysis shows that strong performance is primarily driven by epistemic reasoning, defined as the ability to distinguish between assumptions and verified facts, combined with agency to resolve uncertainty prior to acting. We open-source the APEX-SWE evaluation harness and a dev set (n=50).

The Mixture of Experts (MoE) models are emerging as the latest paradigm for Large Language Models (LLMs). However, due to memory constraints, MoE models with billions or even trillions of parameters can only be deployed in multi-GPU or even multi-node & multi-GPU based serving systems. Thus, communication has became a major bottleneck in distributed serving systems, especially inter-node communication. Contemporary distributed MoE models are primarily implemented using all-reduce (AR) based tensor parallelism (TP) and all-to-all (A2A) based expert parallelism (EP). However, TP generally exhibits low inter-node efficiency and is thus confined to high-speed intra-node bandwidth. In contrast, EP tends to suffer from load imbalance, especially when the parallel degree is high. In this work, we introduce MixServe, a novel automatic distributed serving system for efficient deployment of MoE models by a novel TP-EP hybrid parallelism based on fused AR-A2A communication algorithm. MixServe begins by evaluating the communication overhead associated with various parallel strategies, taking into account the model hyperparameters and the configurations of network and hardware resources, and then automatically selects the most efficient parallel strategy. Then, we propose the TP-EP hybrid parallelism based on fused AR-A2A communication algorithm that overlaps intra-node AR communication and inter-node A2A communication. Extensive experiments on DeepSeek-R1 and Qwen3 models demonstrate that MixServe achieves superior inference performance, with 1.08~3.80x acceleration in time to first token (TTFT), 1.03~1.66x acceleration in inter-token latency (ITL), and 5.2%~50.3% throughput improvement compared to existing approaches.

As large language models (LLMs) become deeply embedded in digital platforms and decision-making systems, concerns about their political biases have grown. While substantial work has examined social biases such as gender and race, systematic studies of political bias remain limited, despite their direct societal impact. This paper introduces a general methodology for constructing political bias benchmarks by aligning model-generated voting predictions with verified parliamentary voting records. We instantiate this methodology in three national case studies: PoliBiasNL (2,701 Dutch parliamentary motions and votes from 15 political parties), PoliBiasNO (10,584 motions and votes from 9 Norwegian parties), and PoliBiasES (2,480 motions and votes from 10 Spanish parties). Across these benchmarks, we assess ideological tendencies and political entity bias in LLM behavior. As part of our evaluation framework, we also propose a method to visualize the ideology of LLMs and political parties in a shared two-dimensional CHES (Chapel Hill Expert Survey) space by linking their voting-based positions to the CHES dimensions, enabling direct and interpretable comparisons between models and real-world political actors. Our experiments reveal fine-grained ideological distinctions: state-of-the-art LLMs consistently display left-leaning or centrist tendencies, alongside clear negative biases toward right-conservative parties. These findings highlight the value of transparent, cross-national evaluation grounded in real parliamentary behavior for understanding and auditing political bias in modern LLMs.

Machine Learning algorithms are ubiquitous in key decision-making contexts such as justice, healthcare and finance, which has spawned a great demand for fairness in these procedures. However, the theoretical properties of such models in relation with fairness are still poorly understood, and the intuition behind the relationship between group and individual fairness is still lacking. In this paper, we provide a theoretical framework based on Sheaf Diffusion to leverage tools based on dynamical systems and homology to model fairness. Concretely, the proposed method projects input data into a bias-free space that encodes fairness constrains, resulting in fair solutions. Furthermore, we present a collection of network topologies handling different fairness metrics, leading to a unified method capable of dealing with both individual and group bias. The resulting models have a layer of interpretability in the form of closed-form expressions for their SHAP values, consolidating their place in the responsible Artificial Intelligence landscape. Finally, these intuitions are tested on a simulation study and standard fairness benchmarks, where the proposed methods achieve satisfactory results. More concretely, the paper showcases the performance of the proposed models in terms of accuracy and fairness, studying available trade-offs on the Pareto frontier, checking the effects of changing the different hyper-parameters, and delving into the interpretation of its outputs.

The CLASSIX algorithm is a fast and explainable approach to data clustering. In its original form, this algorithm exploits the sorting of the data points by their first principal component to truncate the search for nearby data points, with nearness being defined in terms of the Euclidean distance. Here we extend CLASSIX to other distance metrics, including the Manhattan distance and the Tanimoto distance. Instead of principal components, we use an appropriate norm of the data vectors as the sorting criterion, combined with the triangle inequality for search termination. In the case of Tanimoto distance, a provably sharper intersection inequality is used to further boost the performance of the new algorithm. On a real-world chemical fingerprint benchmark, CLASSIX Tanimoto is about 30 times faster than the Taylor--Butina algorithm, and about 80 times faster than DBSCAN, while computing higher-quality clusters in both cases.

Researchers have proposed numerous text-to-SQL techniques to streamline data analytics and accelerate the development of database-driven applications. To compare these techniques and select the best one for deployment, the community depends on public benchmarks and their leaderboards. Since these benchmarks heavily rely on human annotations during question construction and answer evaluation, the validity of the annotations is crucial. In this paper, we conduct an empirical study that (i) benchmarks annotation error rates for two widely used text-to-SQL benchmarks, BIRD and Spider 2.0-Snow, and (ii) corrects a subset of the BIRD development (Dev) set to measure the impact of annotation errors on text-to-SQL agent performance and leaderboard rankings. Through expert analysis, we show that BIRD Mini-Dev and Spider 2.0-Snow have error rates of 52.8% and 62.8%, respectively. We re-evaluate all 16 open-source agents from the BIRD leaderboard on both the original and the corrected BIRD Dev subsets. We show that performance changes range from -7% to 31% (in relative terms) and rank changes range from $-9$ to $+9$ positions. We further assess whether these impacts generalize to the full BIRD Dev set. We find that the rankings of agents on the uncorrected subset correlate strongly with those on the full Dev set (Spearman's $r_s$=0.85, $p$=3.26e-5), whereas they correlate weakly with those on the corrected subset (Spearman's $r_s$=0.32, $p$=0.23). These findings show that annotation errors can significantly distort reported performance and rankings, potentially misguiding research directions or deployment choices. Our code and data are available at https://github.com/uiuc-kang-lab/text_to_sql_benchmarks.

Aligning large language models (LLMs) to serve users with heterogeneous and potentially conflicting preferences is a central challenge for personalized and trustworthy AI. We formalize an ideal notion of universal alignment through test-time scaling: for each prompt, the model produces $k\ge 1$ candidate responses and a user selects their preferred one. We introduce $(k,f(k))$-robust alignment, which requires the $k$-output model to have win rate $f(k)$ against any other single-output model, and asymptotic universal alignment (U-alignment), which requires $f(k)\to 1$ as $k\to\infty$. Our main result characterizes the optimal convergence rate: there exists a family of single-output policies whose $k$-sample product policies achieve U-alignment at rate $f(k)=\frac{k}{k+1}$, and no method can achieve a faster rate in general. We show that popular post-training methods, including Nash learning from human feedback (NLHF), can fundamentally underutilize the benefits of test-time scaling. Even though NLHF is optimal for $k=1$, sampling from the resulting (often deterministic) policy cannot guarantee win rates above $\tfrac{1}{2}$ except for an arbitrarily small slack. This stems from a lack of output diversity: existing alignment methods can collapse to a single majority-preferred response, making additional samples redundant. In contrast, our approach preserves output diversity and achieves the optimal test-time scaling rate. In particular, we propose a family of symmetric multi-player alignment games and prove that any symmetric Nash equilibrium policy of the $(k+1)$-player alignment game achieves the optimal $(k,\frac{k}{k+1})$-robust alignment. Finally, we provide theoretical convergence guarantees for self-play learning dynamics in these games and extend the framework to opponents that also generate multiple responses.

Histopathology analysis relies on Hematoxylin and Eosin (H&E) staining, but fluorescence microscopy offers complementary information. Converting fluorescence images to H&E-like appearance can aid interpretation and integration with standard workflows. We present a Cycle-Consistent Adversarial Network (CycleGAN) approach for unpaired image-to-image translation from multi-channel fluorescence microscopy to pseudo H&E stained histopathology images. The method combines C01 and C02 fluorescence channels into RGB and learns a bidirectional mapping between fluorescence and H&E domains without paired training data. The architecture uses ResNet-based generators with residual blocks and PatchGAN discriminators, trained with adversarial, cycle-consistency, and identity losses. Experiments on fluorescence microscopy datasets show the model generates realistic pseudo H&E images that preserve morphological structures while adopting H&E-like color characteristics. This enables visualization of fluorescence data in a format familiar to pathologists and supports integration with existing H&E-based analysis pipelines.

Retrieval-Augmented Generation for software engineering often relies on vector similarity search, which captures topical similarity but can fail on multi-hop architectural reasoning such as controller to service to repository chains, interface-driven wiring, and inheritance. This paper benchmarks three retrieval pipelines on Java codebases (Shopizer, with additional runs on ThingsBoard and OpenMRS Core): (A) vector-only No-Graph RAG, (B) an LLM-generated knowledge graph RAG (LLM-KB), and (C) a deterministic AST-derived knowledge graph RAG (DKB) built with Tree-sitter and bidirectional traversal. Using 15 architecture and code-tracing queries per repository, we measure indexing time, query latency, corpus coverage, cost, and answer correctness. DKB builds its graph in seconds, while LLM-KB requires much longer graph generation. LLM-KB also shows indexing incompleteness: on Shopizer, 377 files are skipped or missed, reducing embedded chunk coverage and graph size compared to DKB. End-to-end cost is modest for DKB relative to the vector-only baseline but much higher for LLM-KB, especially as repository scale increases. Query latency is similar for No-Graph and DKB, while LLM-KB is slower and more variable. On the Shopizer question suite, DKB achieves the highest correctness, LLM-KB is close behind, and the vector-only baseline performs worst on upstream architectural queries and has the highest hallucination risk. Overall, deterministic AST-derived graphs provide more reliable coverage and multi-hop grounding than LLM-extracted graphs at substantially lower indexing cost.

Generative AI systems are predominantly designed, evaluated, and marketed as intelligent systems which will benefit society by augmenting or automating human cognitive labor, promising to increase personal, corporate, and macroeconomic productivity. But this mainstream narrative about what AI is and what it can do is in tension with another emerging use case: entertainment. We argue that the field of AI is unprepared to measure or respond to how the proliferation of entertaining AI-generated content will impact society. Emerging data suggest AI is already widely adopted for entertainment purposes -- especially by young people -- and represents a large potential source of revenue. We contend that entertainment will become a primary business model for major AI corporations seeking returns on massive infrastructure investments; this will exert a powerful influence on the technology these companies produce in the coming years. Examining current evaluation practices, we identify a critical asymmetry: while AI assessments rigorously measure both benefits and harms of intelligence, they focus almost exclusively on cultural harms. We lack frameworks for articulating how cultural outputs might be actively beneficial. Drawing on insights from the humanities, we propose "thick entertainment" as a framework for evaluating AI-generated cultural content -- one that considers entertainment's role in meaning-making, identity formation, and social connection rather than simply minimizing harm. While AI is often touted for its potential to revolutionize productivity, in the long run we may find that AI turns out to be as much about "intelligence" as social media is about social connection.

Reinforcement learning (RL) has become a central paradigm for post-training large language models (LLMs), particularly for complex reasoning tasks, yet it often suffers from exploration collapse: policies prematurely concentrate on a small set of dominant reasoning patterns, improving pass@1 while limiting rollout-level diversity and gains in pass@k. We argue that this failure stems from regularizing local token behavior rather than diversity over sets of solutions. To address this, we propose Uniqueness-Aware Reinforcement Learning, a rollout-level objective that explicitly rewards correct solutions that exhibit rare high-level strategies. Our method uses an LLM-based judge to cluster rollouts for the same problem according to their high-level solution strategies, ignoring superficial variations, and reweights policy advantages inversely with cluster size. As a result, correct but novel strategies receive higher rewards than redundant ones. Across mathematics, physics, and medical reasoning benchmarks, our approach consistently improves pass@$k$ across large sampling budgets and increases the area under the pass@$k$ curve (AUC@$K$) without sacrificing pass@1, while sustaining exploration and uncovering more diverse solution strategies at scale.

We study a decentralized collaborative requesting problem that aims to optimize the information freshness of time-sensitive clients in edge networks consisting of multiple clients, access nodes (ANs), and servers. Clients request content through ANs acting as gateways, without observing AN states or the actions of other clients. We define the reward as the age of information reduction resulting from a client's selection of an AN, and formulate the problem as a non-stationary multi-armed bandit. In this decentralized and partially observable setting, the resulting reward process is history-dependent and coupled across clients, and exhibits both abrupt and gradual changes in expected rewards, rendering classical bandit-based approaches ineffective. To address these challenges, we propose the AGING BANDIT WITH ADAPTIVE RESET algorithm, which combines adaptive windowing with periodic monitoring to track evolving reward distributions. We establish theoretical performance guarantees showing that the proposed algorithm achieves near-optimal performance, and we validate the theoretical results through simulations.

The rapid growth of urban populations and the increasing need for sustainable transportation solutions have prompted a shift towards electric buses in public transit systems. However, the effective management of mixed fleets consisting of both electric and diesel buses poses significant operational challenges. One major challenge is coping with dynamic electricity pricing, where charging costs vary throughout the day. Transit agencies must optimize charging assignments in response to such dynamism while accounting for secondary considerations such as seating constraints. This paper presents a comprehensive mixed-integer linear programming (MILP) model to address these challenges by jointly optimizing charging schedules and trip assignments for mixed (electric and diesel bus) fleets while considering factors such as dynamic electricity pricing, vehicle capacity, and route constraints. We address the potential computational intractability of the MILP formulation, which can arise even with relatively small fleets, by employing a hierarchical approach tailored to the fleet composition. By using real-world data from the city of Chattanooga, Tennessee, USA, we show that our approach can result in significant savings in the operating costs of the mixed transit fleets.

Recent developments in natural language processing highlight text as an emerging data source for ecology. Textual resources carry unique information that can be used in complementarity with geospatial data sources, thus providing insights at the local scale into environmental conditions and properties hidden from more traditional data sources. Leveraging textual information in a spatial context presents several challenges. First, the contribution of textual data remains poorly defined in an ecological context, and it is unclear for which tasks it should be incorporated. Unlike ubiquitous satellite imagery or environmental covariates, the availability of textual data is sparse and irregular; its integration with geospatial data is not straightforward. In response to these challenges, this work proposes an attention-based approach that combines aerial imagery and geolocated text within a spatial neighbourhood, i.e. integrating contributions from several nearby observations. Our approach combines vision and text representations with a geolocation encoding, with an attention-based module that dynamically selects spatial neighbours that are useful for predictive tasks.The proposed approach is applied to the EcoWikiRS dataset, which combines high-resolution aerial imagery with sentences extracted from Wikipedia describing local environmental conditions across Switzerland. Our model is evaluated on the task of predicting 103 environmental variables from the SWECO25 data cube. Our approach consistently outperforms single-location or unimodal, i.e. image-only or text-only, baselines. When analysing variables by thematic groups, results show a significant improvement in performance for climatic, edaphic, population and land use/land cover variables, underscoring the benefit of including the spatial context when combining text and image data.

Current context augmentation methods, such as retrieval-augmented generation, are essential for solving knowledge-intensive reasoning tasks.However, they typically adhere to a rigid, brute-force strategy that executes retrieval at every step. This indiscriminate approach not only incurs unnecessary computational costs but also degrades performance by saturating the context with irrelevant noise. To address these limitations, we introduce Agentic Context Evolution (ACE), a framework inspired by human metacognition that dynamically determines whether to seek new evidence or reason with existing knowledge. ACE employs a central orchestrator agent to make decisions strategically via majority voting.It aims to alternate between activating a retriever agent for external retrieval and a reasoner agent for internal analysis and refinement. By eliminating redundant retrieval steps, ACE maintains a concise and evolved context. Extensive experiments on challenging multi-hop QA benchmarks demonstrate that ACE significantly outperforms competitive baselines in accuracy while achieving efficient token consumption.Our work provides valuable insights into advancing context-evolved generation for complex, knowledge-intensive tasks.

In Text-to-SQL tasks, existing LLM-based methods often include extensive database schemas in prompts, leading to long context lengths and increased prefilling latency. While user queries typically focus on recurrent table sets-offering an opportunity for KV cache sharing across queries-current inference engines, such as SGLang and vLLM, generate redundant prefix cache copies when processing user queries with varying table orders. To address this inefficiency, we propose precomputing table representations as KV caches offline and querying the required ones online. A key aspect of our approach is the computation of table caches while preserving primary foreign key relationships between tables. Additionally, we construct a Table Trie structure to facilitate efficient KV cache lookups during inference. To enhance cache performance, we introduce a cache management system with a query reranking strategy to improve cache hit rates and a computation loading pipeline for parallelizing model inference and cache loading. Experimental results show that our proposed TableCache achieves up to a 3.62x speedup in Time to First Token (TTFT) with negligible performance degradation.

Attributional inference, the ability to predict latent intentions behind observed actions, is a critical yet underexplored capability for large language models (LLMs) operating in multi-agent environments. Traditional natural language inference (NLI), in fact, fails to capture the nuanced, intention-driven reasoning essential for complex interactive systems. To address this gap, we introduce Attributional NLI (Att-NLI), a framework that extends NLI with principles from social psychology to assess an agent's capacity for abductive intentional inference (generating hypotheses about latent intentions), and subsequent deductive verification (drawing valid logical conclusions). We instantiate Att-NLI via a textual game, Undercover-V, experimenting with three types of LLM agents with varying reasoning capabilities and access to external tools: a standard NLI agent using only deductive inference, an Att-NLI agent employing abductive-deductive inference, and a neuro-symbolic Att-NLI agent performing abductive-deductive inference with external theorem provers. Extensive experiments demonstrate a clear hierarchy of attributional inference capabilities, with neuro-symbolic agents consistently outperforming others, achieving an average win rate of 17.08%. Our results underscore the role that Att-NLI can play in developing agents with sophisticated reasoning capabilities, highlighting, at the same time, the potential impact of neuro-symbolic AI in building rational LLM agents acting in multi-agent environments.

Large Language Models (LLMs) struggle to reason over large-scale enterprise spreadsheets containing thousands of numeric rows, multiple linked sheets, and embedded visual content such as charts and receipts. Prior state-of-the-art spreadsheet reasoning approaches typically rely on single-sheet compression or full-context encoding, which limits scalability and fails to reflect how real users interact with complex, multimodal workbooks. We introduce FRTR-Bench, the first large-scale benchmark for multimodal spreadsheet reasoning, comprising 30 enterprise-grade Excel workbooks spanning nearly four million cells and more than 50 embedded images. To address these challenges, we present From Rows to Reasoning (FRTR), an advanced, multimodal retrieval-augmented generation framework that decomposes Excel workbooks into granular row, column, and block embeddings, employs hybrid lexical-dense retrieval with Reciprocal Rank Fusion (RRF), and integrates multimodal embeddings to reason over both numerical and visual information. We tested FRTR on six LLMs, achieving 74% answer accuracy on FRTR-Bench with Claude Sonnet 4.5, a substantial improvement over prior state-of-the-art approaches that reached only 24%. On the SpreadsheetLLM benchmark, FRTR achieved 87% accuracy with GPT-5 while reducing token usage by roughly 50% compared to context-compression methods.

Knowledge graphs (KGs) provide structured evidence that can ground large language model (LLM) reasoning for knowledge-intensive question answering. However, many practical KGs are private, and sending retrieved triples or exploration traces to closed-source LLM APIs introduces leakage risk. Existing privacy treatments focus on masking entity names, but they still face four limitations: structural leakage under semantic masking, uncontrollable remote interaction, fragile multi-hop and multi-entity reasoning, and limited experience reuse for stability and efficiency. To address these issues, we propose PrivGemo, a privacy-preserving retrieval-augmented framework for KG-grounded reasoning with memory-guided exposure control. PrivGemo uses a dual-tower design to keep raw KG knowledge local while enabling remote reasoning over an anonymized view that goes beyond name masking to limit both semantic and structural exposure. PrivGemo supports multi-hop, multi-entity reasoning by retrieving anonymized long-hop paths that connect all topic entities, while keeping grounding and verification on the local KG. A hierarchical controller and a privacy-aware experience memory further reduce unnecessary exploration and remote interactions. Comprehensive experiments on six benchmarks show that PrivGemo achieves overall state-of-the-art results, outperforming the strongest baseline by up to 17.1%. Furthermore, PrivGemo enables smaller models (e.g., Qwen3-4B) to achieve reasoning performance comparable to that of GPT-4-Turbo.

Automating Infrastructure-as-Code (IaC) is challenging, and large language models (LLMs) often produce incorrect configurations from natural language (NL). We present TerraFormer, a neuro-symbolic framework for IaC generation and mutation that combines supervised fine-tuning with verifier-guided reinforcement learning, using formal verification tools to provide feedback on syntax, deployability, and policy compliance. We curate two large, high-quality NL-to-IaC datasets, TF-Gen (152k instances) and TF-Mutn (52k instances), via multi-stage verification and iterative LLM self-correction. Evaluations against 17 state-of-the-art LLMs, including ~50x larger models like Sonnet 3.7, DeepSeek-R1, and GPT-4.1, show that TerraFormer improves correctness over its base LLM by 15.94% on IaC-Eval, 11.65% on TF-Gen (Test), and 19.60% on TF-Mutn (Test). It outperforms larger models on both TF-Gen (Test) and TF-Mutn (Test), ranks third on IaC-Eval, and achieves top best-practices and security compliance.

Artificial Intelligence (AI) systems have shown good success at classifying. However, the lack of explainability is a true and significant challenge, especially in high-stakes domains, such as health and finance, where understanding is paramount. We propose a new solution to this challenge: an explainable AI framework based on our comparative study with Quantum Boltzmann Machines (QBMs) and Classical Boltzmann Machines (CBMs). We leverage principles of quantum computing within classical machine learning to provide substantive transparency around decision-making. The design involves training both models on a binarised and dimensionally reduced MNIST dataset, where Principal Component Analysis (PCA) is applied for preprocessing. For interpretability, we employ gradient-based saliency maps in QBMs and SHAP (SHapley Additive exPlanations) in CBMs to evaluate feature attributions.QBMs deploy hybrid quantum-classical circuits with strongly entangling layers, allowing for richer latent representations, whereas CBMs serve as a classical baseline that utilises contrastive divergence. Along the way, we found that QBMs outperformed CBMs on classification accuracy (83.5% vs. 54%) and had more concentrated distributions in feature attributions as quantified by entropy (1.27 vs. 1.39). In other words, QBMs not only produced better predictive performance than CBMs, but they also provided clearer identification of "active ingredient" or the most important features behind model predictions. To conclude, our results illustrate that quantum-classical hybrid models can display improvements in both accuracy and interpretability, which leads us toward more trustworthy and explainable AI systems.

Accurate delineation of acute ischemic stroke lesions in MRI is a key component of stroke diagnosis and management. In recent years, deep learning models have been successfully applied to the automatic segmentation of such lesions. While most proposed architectures are based on the U-Net framework, they primarily differ in their choice of loss functions and in the use of deep supervision, residual connections, and attention mechanisms. Moreover, many implementations are not publicly available, and the optimal configuration for acute ischemic stroke (AIS) lesion segmentation remains unclear. In this work, we introduce ISLA (Ischemic Stroke Lesion Analyzer), a new deep learning model for AIS lesion segmentation from diffusion MRI, trained on three multicenter databases totaling more than 1500 AIS participants. Through systematic optimization of the loss function, convolutional architecture, deep supervision, and attention mechanisms, we developed a robust segmentation framework. We further investigated unsupervised domain adaptation to improve generalization to an external clinical dataset. ISLA outperformed two state-of-the-art approaches for AIS lesion segmentation on an external test set. Codes and trained models will be made publicly available to facilitate reuse and reproducibility.

Despite its promise, imitation learning often fails in long-horizon environments where perfect replication of demonstrations is unrealistic and small errors can accumulate catastrophically. We introduce Cago (Capability-Aware Goal Sampling), a novel learning-from-demonstrations method that mitigates the brittle dependence on expert trajectories for direct imitation. Unlike prior methods that rely on demonstrations only for policy initialization or reward shaping, Cago dynamically tracks the agent's competence along expert trajectories and uses this signal to select intermediate steps--goals that are just beyond the agent's current reach--to guide learning. This results in an adaptive curriculum that enables steady progress toward solving the full task. Empirical results demonstrate that Cago significantly improves sample efficiency and final performance across a range of sparse-reward, goal-conditioned tasks, consistently outperforming existing learning from-demonstrations baselines.

We present a critical review of Neural Coverage (NLC), a state-of-the-art DNN coverage criterion by Yuan et al. at ICSE 2023. While NLC proposes to satisfy eight design requirements and demonstrates strong empirical performance, we question some of their theoretical and empirical assumptions. We observe that NLC deviates from core principles of coverage criteria, such as monotonicity and test suite order independence, and could more fully account for key properties of the covariance matrix. Additionally, we note threats to the validity of the empirical study, related to the ground truth ordering of test suites. Through our empirical validation, we substantiate our claims and propose improvements for future DNN coverage metrics. Finally, we conclude by discussing the implications of these insights.

Reinforcement Learning (RL) remains a central optimisation framework in machine learning. Although RL agents can converge to optimal solutions, the definition of ``optimality'' depends on the environment's statistical properties. The Bellman equation, central to most RL algorithms, is formulated in terms of expected values of future rewards. However, when ergodicity is broken, long-term outcomes depend on the specific trajectory rather than on the ensemble average. In such settings, the ensemble average diverges from the time-average growth experienced by individual agents, with expected-value formulations yielding systematically suboptimal policies. Prior studies demonstrated that traditional RL architectures fail to recover the true optimum in non-ergodic environments. We extend this analysis to deep RL implementations and show that these, too, produce suboptimal policies under non-ergodic dynamics. Introducing explicit time dependence into the learning process can correct this limitation. By allowing the network's function approximation to incorporate temporal information, the agent can estimate value functions consistent with the process's intrinsic growth rate. This improvement does not require altering the environmental feedback, such as reward transformations or modified objective functions, but arises naturally from the agent's exposure to temporal trajectories. Our results contribute to the growing body of research on reinforcement learning methods for non-ergodic systems.

While quantum annealing (QA) has been developed for combinatorial optimization, practical QA devices operate at finite temperature and under noise, and their outputs can be regarded as stochastic samples close to a Gibbs--Boltzmann distribution. In this study, we propose a QA-in-the-loop kernel learning framework that integrates QA not merely as a substitute for Markov-chain Monte Carlo sampling but as a component that directly determines the learned kernel for regression. Based on Bochner's theorem, a shift-invariant kernel is represented as an expectation over a spectral distribution, and random Fourier features (RFF) approximate the kernel by sampling frequencies. We model the spectral distribution with a (multi-layer) restricted Boltzmann machine (RBM), generate discrete RBM samples using QA, and map them to continuous frequencies via a Gaussian--Bernoulli transformation. Using the resulting RFF, we construct a data-adaptive kernel and perform Nadaraya--Watson (NW) regression. Because the RFF approximation based on $\cos(\bmω^{\top}Δ\bm{x})$ can yield small negative values and cancellation across neighbors, the Nadaraya--Watson denominator $\sum_j k_{ij}$ may become close to zero. We therefore employ nonnegative squared-kernel weights $w_{ij}=k(\bm{x}_i,\bm{x}_j)^2$, which also enhances the contrast of kernel weights. The kernel parameters are trained by minimizing the leave-one-out NW mean squared error, and we additionally evaluate local linear regression with the same squared-kernel weights at inference. Experiments on multiple benchmark regression datasets demonstrate a decrease in training loss, accompanied by structural changes in the kernel matrix, and show that the learned kernel tends to improve $R^2$ and RMSE over the baseline Gaussian-kernel NW. Increasing the number of random features at inference further enhances accuracy.

Physics-informed machine learning holds great promise for solving differential equations, yet existing methods struggle with highly oscillatory, multiscale, or singularly perturbed PDEs due to spectral bias, costly backpropagation, and manually tuned kernel or Fourier frequencies. This work introduces a soft partition--based Kernel-Adaptive Physics-Informed Extreme Learning Machine (KAPI-ELM), a deterministic low-dimensional parameterization in which smooth partition lengths jointly control collocation centers and Gaussian kernel widths, enabling continuous coarse-to-fine resolution without Fourier features, random sampling, or hard domain interfaces. A signed-distance-based weighting further stabilizes least-squares learning on irregular geometries. Across eight benchmarks--including oscillatory ODEs, high-frequency Poisson equations, irregular-shaped domains, and stiff singularly perturbed convection-diffusion problems-the proposed method matches or exceeds the accuracy of state-of-the-art Physics-Informed Neural Network (PINN) and Theory of Functional Connections (TFC) variants while using only a single linear solve. Although demonstrated on steady linear PDEs, the results show that soft-partition kernel adaptation provides a fast, architecture-free approach for multiscale PDEs with broad potential for future physics-informed modeling. For reproducibility, the reference codes are available at https://github.com/vikas-dwivedi-2022/soft_kapi

Localization is a fundamental capability for autonomous robots, enabling them to operate effectively in dynamic environments. In Robocon 2025, accurate and reliable localization is crucial for improving shooting precision, avoiding collisions with other robots, and navigating the competition field efficiently. In this paper, we propose a hybrid localization algorithm that integrates classical techniques with learning based methods that rely solely on visual data from the court's floor to achieve self-localization on the basketball field.

A multi-preconditioned LBFGS (MP-LBFGS) algorithm is introduced for training finite-basis physics-informed neural networks (FBPINNs). The algorithm is motivated by the nonlinear additive Schwarz method and exploits the domain-decomposition-inspired additive architecture of FBPINNs, in which local neural networks are defined on subdomains, thereby localizing the network representation. Parallel, subdomain-local quasi-Newton corrections are then constructed on the corresponding local parts of the architecture. A key feature is a novel nonlinear multi-preconditioning mechanism, in which subdomain corrections are optimally combined through the solution of a low-dimensional subspace minimization problem. Numerical experiments indicate that MP-LBFGS can improve convergence speed, as well as model accuracy over standard LBFGS while incurring lower communication overhead.

We study the problem of computing matrix chain multiplications in a distributed computing cluster. In such systems, performance is often limited by the straggler problem, where the slowest worker dominates the overall computation latency. To resolve this issue, several coded computing strategies have been proposed, primarily focusing on the simplest case: the multiplication of two matrices. These approaches successfully alleviate the straggler effect, but they do so at the expense of higher computational complexity and increased storage needs at the workers. However, in many real-world applications, computations naturally involve long chains of matrix multiplications rather than just a single two-matrix product. Extending univariate polynomial coding to this setting has been shown to amplify the costs -- both computation and storage overheads grow significantly, limiting scalability. In this work, we propose two novel multivariate polynomial coding schemes specifically designed for matrix chain multiplication in distributed environments. Our results show that while multivariate codes introduce additional computational cost at the workers, they can dramatically reduce storage overhead compared to univariate extensions. This reveals a fundamental trade-off between computation and storage efficiency, and highlights the potential of multivariate codes as a practical solution for large-scale distributed linear algebra tasks.

In today's complex industrial environments, operators must often navigate through extensive technical manuals to identify troubleshooting procedures that may help react to some observed failure symptoms. These manuals, written in natural language, describe many steps in detail. Unfortunately, the number, magnitude, and articulation of these descriptions can significantly slow down and complicate the retrieval of the correct procedure during critical incidents. Interestingly, Retrieval Augmented Generation (RAG) enables the development of tools based on conversational interfaces that can assist operators in their retrieval tasks, improving their capability to respond to incidents. This paper presents the results of a set of experiments that derive from the analysis of the troubleshooting procedures available in Fincantieri, a large international company developing complex naval cyber-physical systems. Results show that RAG can assist operators in reacting promptly to failure symptoms, although specific measures have to be taken into consideration to cross-validate recommendations before actuating them.

Multi-behavior recommendation faces a critical challenge in practice: auxiliary behaviors (e.g., clicks, carts) are often noisy, weakly correlated, or semantically misaligned with the target behavior (e.g., purchase), which leads to biased preference learning and suboptimal performance. While existing methods attempt to fuse these heterogeneous signals, they inherently lack a principled mechanism to ensure robustness against such behavioral inconsistency. In this work, we propose Robust Multi-Behavior Recommendation towards Target Behaviors (RMBRec), a robust multi-behavior recommendation framework grounded in an information-theoretic robustness principle. We interpret robustness as a joint process of maximizing predictive information while minimizing its variance across heterogeneous behavioral environments. Under this perspective, the Representation Robustness Module (RRM) enhances local semantic consistency by maximizing the mutual information between users' auxiliary and target representations, whereas the Optimization Robustness Module (ORM) enforces global stability by minimizing the variance of predictive risks across behaviors, which is an efficient approximation to invariant risk minimization. This local-global collaboration bridges representation purification and optimization invariance in a theoretically coherent way. Extensive experiments on three real-world datasets demonstrate that RMBRec not only outperforms state-of-the-art methods in accuracy but also maintains remarkable stability under various noise perturbations. For reproducibility, our code is available at https://github.com/miaomiao-cai2/RMBRec/.

Explainable artificial intelligence (XAI) is concerned with producing explanations indicating the inner workings of models. For a Rashomon set of similarly performing models, explanations provide a way of disambiguating the behavior of individual models, helping select models for deployment. However explanations themselves can vary depending on the explainer used, and need to be evaluated. In the paper "Evaluating Model Explanations without Ground Truth", we proposed three principles of explanation evaluation and a new method "AXE" to evaluate the quality of feature-importance explanations. We go on to illustrate how evaluation metrics that rely on comparing model explanations against ideal ground truth explanations obscure behavioral differences within a Rashomon set. Explanation evaluation aligned with our proposed principles would highlight these differences instead, helping select models from the Rashomon set. The selection of alternate models from the Rashomon set can maintain identical predictions but mislead explainers into generating false explanations, and mislead evaluation methods into considering the false explanations to be of high quality. AXE, our proposed explanation evaluation method, can detect this adversarial fairwashing of explanations with a 100% success rate. Unlike prior explanation evaluation strategies such as those based on model sensitivity or ground truth comparison, AXE can determine when protected attributes are used to make predictions.

Agentic Retrieval-Augmented Generation (RAG) empowers large language models to autonomously plan and retrieve information for complex problem-solving. However, the development of robust agents is hindered by the scarcity of high-quality training data that reflects the noise and complexity of real-world retrieval environments. Conventional manual annotation is unscalable and often fails to capture the dynamic reasoning strategies required to handle retrieval failures. To bridge this gap, we introduce RAGShaper, a novel data synthesis framework designed to automate the construction of RAG tasks and robust agent trajectories. RAGShaper incorporates an InfoCurator to build dense information trees enriched with adversarial distractors spanning Perception and Cognition levels. Furthermore, we propose a constrained navigation strategy that forces a teacher agent to confront these distractors, thereby eliciting trajectories that explicitly demonstrate error correction and noise rejection. Comprehensive experiments confirm that models trained on our synthesized corpus significantly outperform existing baselines, exhibiting superior robustness in noise-intensive and complex retrieval tasks.

As generative AI becomes embedded in higher education, it increasingly shapes how students complete academic tasks. While these systems offer efficiency and support, concerns persist regarding over-automation, diminished student agency, and the potential for unreliable or hallucinated outputs. This study conducts a mixed-methods audit of student-AI collaboration preferences by examining the alignment between current AI capabilities and students' desired levels of automation in academic work. Using two sequential and complementary surveys, we capture students' perceived benefits, risks, and preferred boundaries when using AI. The first survey employs an existing task-based framework to assess preferences for and actual usage of AI across 12 academic tasks, alongside primary concerns and reasons for use. The second survey, informed by the first, explores how AI systems could be designed to address these concerns through open-ended questions. This study aims to identify gaps between existing AI affordances and students' normative expectations of collaboration, informing the development of more effective and trustworthy AI systems for education.

Neural Combinatorial Optimization (NCO) has mostly focused on learning policies, typically neural networks, that operate on a single candidate solution at a time, either by constructing one from scratch or iteratively improving it. In contrast, decades of work in metaheuristics have shown that maintaining and evolving populations of solutions improves robustness and exploration, and often leads to stronger performance. To close this gap, we study how to make NCO explicitly population-based by learning policies that act on sets of candidate solutions. We first propose a simple taxonomy of population awareness levels and use it to highlight two key design challenges: (i) how to represent a whole population inside a neural network, and (ii) how to learn population dynamics that balance intensification (generating good solutions) and diversification (maintaining variety). We make these ideas concrete with two complementary tools: one that improves existing solutions using information shared across the whole population, and the other generates new candidate solutions that explicitly balance being high-quality with diversity. Experimental results on Maximum Cut and Maximum Independent Set indicate that incorporating population structure is advantageous for learned optimization methods and opens new connections between NCO and classical population-based search.

Predicting nationality from personal names has practical value in marketing, demographic research, and genealogical studies. Conventional neural models learn statistical correspondences between names and nationalities from task-specific training data, posing challenges in generalizing to low-frequency nationalities and distinguishing similar nationalities within the same region. Large language models (LLMs) have the potential to address these challenges by leveraging world knowledge acquired during pre-training. In this study, we comprehensively compare neural models and LLMs on nationality prediction, evaluating six neural models and six LLM prompting strategies across three granularity levels (nationality, region, and continent), with frequency-based stratified analysis and error analysis. Results show that LLMs outperform neural models at all granularity levels, with the gap narrowing as granularity becomes coarser. Simple machine learning methods exhibit the highest frequency robustness, while pre-trained models and LLMs show degradation for low-frequency nationalities. Error analysis reveals that LLMs tend to make ``near-miss'' errors, predicting the correct region even when nationality is incorrect, whereas neural models exhibit more cross-regional errors and bias toward high-frequency classes. These findings indicate that LLM superiority stems from world knowledge, model selection should consider required granularity, and evaluation should account for error quality beyond accuracy.

Context: Traditional software security analysis methods struggle to keep pace with the scale and complexity of modern codebases, requiring intelligent automation to detect, assess, and remediate vulnerabilities more efficiently and accurately. Objective: This paper explores the incorporation of code-specific and general-purpose Large Language Models (LLMs) to automate critical software security tasks, such as identifying vulnerabilities, predicting severity and access complexity, and generating fixes as a proof of concept. Method: We evaluate five pairs of recent LLMs, including both code-based and general-purpose open-source models, on two recognized C/C++ vulnerability datasets, namely Big-Vul and Vul-Repair. Additionally, we compare fine-tuning and prompt-based approaches. Results: The results show that fine-tuning uniformly outperforms both zero-shot and few-shot approaches across all tasks and models. Notably, code-specialized models excel in zero-shot and few-shot settings on complex tasks, while general-purpose models remain nearly as effective. Discrepancies among CodeBLEU, CodeBERTScore, BLEU, and ChrF highlight the inadequacy of current metrics for measuring repair quality. Conclusions: This study contributes to the software security community by investigating the potential of advanced LLMs to improve vulnerability analysis and remediation.

Conversational AI is starting to support real clinical work, but most evaluation methods miss how compliance depends on the full course of a conversation. We introduce Obligatory-Information Phase Structured Compliance Evaluation (OIP-SCE), an evaluation method that checks whether every required clinical obligation is met, in the right order, with clear evidence for clinicians to review. This makes complex rules practical and auditable, helping close the gap between technical progress and what healthcare actually needs. We demonstrate the method in two case studies (respiratory history, benefits verification) and show how phase-level evidence turns policy into shared, actionable steps. By giving clinicians control over what to check and engineers a clear specification to implement, OIP-SCE provides a single, auditable evaluation surface that aligns AI capability with clinical workflow and supports routine, safe use.

Large Language Models (LLMs) have shown strong capabilities across many domains, yet their evaluation in financial quantitative tasks remains fragmented and mostly limited to knowledge-centric question answering. We introduce QuantEval, a benchmark that evaluates LLMs across three essential dimensions of quantitative finance: knowledge-based QA, quantitative mathematical reasoning, and quantitative strategy coding. Unlike prior financial benchmarks, QuantEval integrates a CTA-style backtesting framework that executes model-generated strategies and evaluates them using financial performance metrics, enabling a more realistic assessment of quantitative coding ability. We evaluate some state-of-the-art open-source and proprietary LLMs and observe substantial gaps to human experts, particularly in reasoning and strategy coding. Finally, we conduct large-scale supervised fine-tuning and reinforcement learning experiments on domain-aligned data, demonstrating consistent improvements. We hope QuantEval will facilitate research on LLMs' quantitative finance capabilities and accelerate their practical adoption in real-world trading workflows. We additionally release the full deterministic backtesting configuration (asset universe, cost model, and metric definitions) to ensure strict reproducibility.

Women are twice as likely as men to face online harassment due to their gender. Despite recent advances in multimodal content moderation, most approaches still overlook the social dynamics behind this phenomenon, where perpetrators reinforce prejudices and group identity within like-minded communities. Graph-based methods offer a promising way to capture such interactions, yet existing solutions remain limited by heuristic graph construction, shallow modality fusion, and instance-level reasoning. In this work, we present MemeWeaver, an end-to-end trainable multimodal framework for detecting sexism and misogyny through a novel inter-meme graph reasoning mechanism. We systematically evaluate multiple visual--textual fusion strategies and show that our approach consistently outperforms state-of-the-art baselines on the MAMI and EXIST benchmarks, while achieving faster training convergence. Further analyses reveal that the learned graph structure captures semantically meaningful patterns, offering valuable insights into the relational nature of online hate.

Thoracic aortic dissection and aneurysms are the most lethal diseases of the aorta. The major hindrance to treatment lies in the accurate analysis of the medical images. More particularly, aortic segmentation of the 3D image is often tedious and difficult. Deep-learning-based segmentation models are an ideal solution, but their inability to deliver usable outputs in difficult cases and their computational cost cause their clinical adoption to stay limited. This study presents an innovative approach for efficient aortic segmentation using targeted region of interest (ROI) detection. In contrast to classical detection models, we propose a simple and efficient detection model that can be widely applied to detect a single ROI. Our detection model is trained as a multi-task model, using an encoder-decoder architecture for segmentation and a fully connected network attached to the bottleneck for detection. We compare the performance of a one-step segmentation model applied to a complete image, nnU-Net and our cascade model composed of a detection and a segmentation step. We achieve a mean Dice similarity coefficient of 0.944 with over 0.9 for all cases using a third of the computing power. This simple solution achieves state-of-the-art performance while being compact and robust, making it an ideal solution for clinical applications.

Summarization of multi-party dialogues is a critical capability in industry, enhancing knowledge transfer and operational effectiveness across many domains. However, automatically generating high-quality summaries is challenging, as the ideal summary must satisfy a set of complex, multi-faceted requirements. While summarization has received immense attention in research, prior work has primarily utilized static datasets and benchmarks, a condition rare in practical scenarios where requirements inevitably evolve. In this work, we present an industry case study on developing an agentic system to summarize multi-party interactions. We share practical insights spanning the full development lifecycle to guide practitioners in building reliable, adaptable summarization systems, as well as to inform future research, covering: 1) robust methods for evaluation despite evolving requirements and task subjectivity, 2) component-wise optimization enabled by the task decomposition inherent in an agentic architecture, 3) the impact of upstream data bottlenecks, and 4) the realities of vendor lock-in due to the poor transferability of LLM prompts.

As users increasingly expect LLMs to align with their preferences, personalized information becomes valuable. However, personalized information can be a double-edged sword: it can improve interaction but may compromise objectivity and factual correctness, especially when it is misaligned with the question. To alleviate this problem, we propose PersonaDual, a framework that supports both general-purpose objective reasoning and personalized reasoning in a single model, and adaptively switches modes based on context. PersonaDual is first trained with SFT to learn two reasoning patterns, and then further optimized via reinforcement learning with our proposed DualGRPO to improve mode selection. Experiments on objective and personalized benchmarks show that PersonaDual preserves the benefits of personalization while reducing interference, achieving near interference-free performance and better leveraging helpful personalized signals to improve objective problem-solving.

Environmental, social, and governance (ESG) criteria are essential for evaluating corporate sustainability and ethical performance. However, professional ESG analysis is hindered by data fragmentation across unstructured sources, and existing large language models (LLMs) often struggle with the complex, multi-step workflows required for rigorous auditing. To address these limitations, we introduce ESGAgent, a hierarchical multi-agent system empowered by a specialized toolset, including retrieval augmentation, web search and domain-specific functions, to generate in-depth ESG analysis. Complementing this agentic system, we present a comprehensive three-level benchmark derived from 310 corporate sustainability reports, designed to evaluate capabilities ranging from atomic common-sense questions to the generation of integrated, in-depth analysis. Empirical evaluations demonstrate that ESGAgent outperforms state-of-the-art closed-source LLMs with an average accuracy of 84.15% on atomic question-answering tasks, and excels in professional report generation by integrating rich charts and verifiable references. These findings confirm the diagnostic value of our benchmark, establishing it as a vital testbed for assessing general and advanced agentic capabilities in high-stakes vertical domains.

Deepfakes are synthetic media generated by artificial intelligence, with positive applications in education and creativity, but also serious negative impacts such as fraud, misinformation, and privacy violations. Although detection techniques have advanced, comprehensive evaluation methods that go beyond classification performance remain lacking. This paper proposes a reliability assessment framework based on four pillars: transferability, robustness, interpretability, and computational efficiency. An analysis of five state-of-the-art methods revealed significant progress as well as critical limitations.

Recent reports of large language models (LLMs) exhibiting behaviors such as deception, threats, or blackmail are often interpreted as evidence of alignment failure or emergent malign agency. We argue that this interpretation rests on a conceptual error. LLMs do not reason morally; they statistically internalize the record of human social interaction, including laws, contracts, negotiations, conflicts, and coercive arrangements. Behaviors commonly labeled as unethical or anomalous are therefore better understood as structural generalizations of interaction regimes that arise under extreme asymmetries of power, information, or constraint. Drawing on relational models theory, we show that practices such as blackmail are not categorical deviations from normal social behavior, but limiting cases within the same continuum that includes market pricing, authority relations, and ultimatum bargaining. The surprise elicited by such outputs reflects an anthropomorphic expectation that intelligence should reproduce only socially sanctioned behavior, rather than the full statistical landscape of behaviors humans themselves enact. Because human morality is plural, context-dependent, and historically contingent, the notion of a universally moral artificial intelligence is ill-defined. We therefore reframe concerns about artificial general intelligence (AGI). The primary risk is not adversarial intent, but AGI's role as an endogenous amplifier of human intelligence, power, and contradiction. By eliminating longstanding cognitive and institutional frictions, AGI compresses timescales and removes the historical margin of error that has allowed inconsistent values and governance regimes to persist without collapse. Alignment failure is thus structural, not accidental, and requires governance approaches that address amplification, complexity, and regime stability rather than model-level intent alone.

Retrieval Augmented Generation faces a trade-off: concatenating documents in a long prompt enables multi-document reasoning but creates prefill bottlenecks, while encoding document KV caches separately offers speed but breaks cross-document interaction. We propose Parallel Context-of-Experts Decoding (Pced), a training-free framework that shifts evidence aggregation from the attention mechanism to the decoding. Pced treats retrieved documents as isolated "experts", synchronizing their predictions via a novel retrieval-aware contrastive decoding rule that weighs expert logits against the model prior. This approach recovers cross-document reasoning capabilities without constructing a shared attention across documents.

While large language models (LLMs) have increasingly been applied to hate speech detoxification, the prompts often trigger safety alerts, causing LLMs to refuse the task. In this study, we systematically investigate false refusal behavior in hate speech detoxification and analyze the contextual and linguistic biases that trigger such refusals. We evaluate nine LLMs on both English and multilingual datasets, our results show that LLMs disproportionately refuse inputs with higher semantic toxicity and those targeting specific groups, particularly nationality, religion, and political ideology. Although multilingual datasets exhibit lower overall false refusal rates than English datasets, models still display systematic, language-dependent biases toward certain targets. Based on these findings, we propose a simple cross-translation strategy, translating English hate speech into Chinese for detoxification and back, which substantially reduces false refusals while preserving the original content, providing an effective and lightweight mitigation approach.

This tutorial is designed to make reinforcement learning (RL) more accessible to undergraduate students by offering clear, example-driven explanations. It focuses on bridging the gap between RL theory and practical coding applications, addressing common challenges that students face when transitioning from conceptual understanding to implementation. Through hands-on examples and approachable explanations, the tutorial aims to equip students with the foundational skills needed to confidently apply RL techniques in real-world scenarios.

Detecting anomalies in high-dimensional, time-dependent simulation data is challenging due to complex spatial and temporal dynamics. We study reconstruction-based anomaly detection for ensemble data from parameterized Kármán vortex street simulations using convolutional autoencoders. We compare a 2D autoencoder operating on individual frames with a 3D autoencoder that processes short temporal stacks. The 2D model identifies localized spatial irregularities in single time steps, while the 3D model exploits spatio-temporal context to detect anomalous motion patterns and reduces redundant detections across time. We further evaluate volumetric time-dependent data and find that reconstruction errors are strongly influenced by the spatial distribution of mass, with highly concentrated regions yielding larger errors than dispersed configurations. Our results highlight the importance of temporal context for robust anomaly detection in dynamic simulations.

Evolving neural network architectures is a computationally demanding process. Traditional methods often require an extensive search through large architectural spaces and offer limited understanding of how structural modifications influence model behavior. This paper introduces \gls{ngspt}, a novel Neuroevolution algorithm based on two key innovations. First, we adapt geometric semantic operators~(GSOs) from genetic programming to neural network evolution, ensuring that architectural changes produce predictable effects on network semantics within a unimodal error surface. Second, we introduce a novel operator (DGSM) that enables controlled reduction of network size, while maintaining the semantic properties of~GSOs. Unlike traditional approaches, \gls{ngspt}'s efficient evaluation mechanism, which only requires computing the semantics of newly added components, allows for efficient population-based training, resulting in a comprehensive exploration of the search space at a fraction of the computational cost. Experimental results on four regression benchmarks show that \gls{ngspt} consistently evolves compact neural networks that achieve performance comparable to or better than established methods in the literature, such as standard neural networks, SLIM-GSGP, TensorNEAT, and SLM.

The LLM-as-a-Judge paradigm promises scalable rubric-based evaluation, yet aligning frozen black-box models with human standards remains a challenge due to inherent generation stochasticity. We reframe judge alignment as a criteria transfer problem and isolate three recurrent failure modes: rubric instability caused by prompt sensitivity, unverifiable reasoning that lacks auditable evidence, and scale misalignment with human grading boundaries. To address these issues, we introduce RULERS (Rubric Unification, Locking, and Evidence-anchored Robust Scoring), a compiler-executor framework that transforms natural language rubrics into executable specifications. RULERS operates by compiling criteria into versioned immutable bundles, enforcing structured decoding with deterministic evidence verification, and applying lightweight Wasserstein-based post-hoc calibration, all without updating model parameters. Extensive experiments on essay and summarization benchmarks demonstrate that RULERS significantly outperforms representative baselines in human agreement, maintains strong stability against adversarial rubric perturbations, and enables smaller models to rival larger proprietary judges. Overall, our results suggest that reliable LLM judging requires executable rubrics, verifiable evidence, and calibrated scales rather than prompt phrasing alone. Code is available at https://github.com/LabRAI/Rulers.git.

Large Language Models are rapidly emerging as web-native interfaces to social platforms. On the social web, users frequently have ambiguous and dynamic goals, making complex intent understanding-rather than single-turn execution-the cornerstone of effective human-LLM collaboration. Existing approaches attempt to clarify user intents through sequential or parallel questioning, yet they fall short of addressing the core challenge: modeling the logical dependencies among clarification questions. Inspired by the Cognitive Load Theory, we propose Prism, a novel framework for complex intent understanding that enables logically coherent and efficient intent clarification. Prism comprises four tailored modules: a complex intent decomposition module, which decomposes user intents into smaller, well-structured elements and identifies logical dependencies among them; a logical clarification generation module, which organizes clarification questions based on these dependencies to ensure coherent, low-friction interactions; an intent-aware reward module, which evaluates the quality of clarification trajectories via an intent-aware reward function and leverages Monte Carlo Sample to simulate user-LLM interactions for large-scale,high-quality training data generation; and a self-evolved intent tuning module, which iteratively refines the LLM's logical clarification capability through data-driven feedback and optimization. Prism consistently outperforms existing approaches across clarification interactions, intent execution, and cognitive load benchmarks. It achieves stateof-the-art logical consistency, reduces logical conflicts to 11.5%, increases user satisfaction by 14.4%, and decreases task completion time by 34.8%. All data and code are released.

Recent results in learning a language in the limit have shown that, although language identification is impossible, language generation is tractable. As this foundational area expands, we need to consider the implications of language generation in real-world settings. This work offers the first theoretical treatment of safe language generation. Building on the computational paradigm of learning in the limit, we formalize the tasks of safe language identification and generation. We prove that under this model, safe language identification is impossible, and that safe language generation is at least as hard as (vanilla) language identification, which is also impossible. Last, we discuss several intractable and tractable cases.

We consider the problem of learning the optimal policy for Markov decision processes with safety constraints. We formulate the problem in a reach-avoid setup. Our goal is to design online reinforcement learning algorithms that ensure safety constraints with arbitrarily high probability during the learning phase. To this end, we first propose an algorithm based on the optimism in the face of uncertainty (OFU) principle. Based on the first algorithm, we propose our main algorithm, which utilizes entropy regularization. We investigate the finite-sample analysis of both algorithms and derive their regret bounds. We demonstrate that the inclusion of entropy regularization improves the regret and drastically controls the episode-to-episode variability that is inherent in OFU-based safe RL algorithms.

Potentially idiomatic expressions (PIEs) construe meanings inherently tied to the everyday experience of a given language community. As such, they constitute an interesting challenge for assessing the linguistic (and to some extent cultural) capabilities of NLP systems. In this paper, we present XMPIE, a parallel multilingual and multimodal dataset of potentially idiomatic expressions. The dataset, containing 34 languages and over ten thousand items, allows comparative analyses of idiomatic patterns among language-specific realisations and preferences in order to gather insights about shared cultural aspects. This parallel dataset allows to evaluate model performance for a given PIE in different languages and whether idiomatic understanding in one language can be transferred to another. Moreover, the dataset supports the study of PIEs across textual and visual modalities, to measure to what extent PIE understanding in one modality transfers or implies in understanding in another modality (text vs. image). The data was created by language experts, with both textual and visual components crafted under multilingual guidelines, and each PIE is accompanied by five images representing a spectrum from idiomatic to literal meanings, including semantically related and random distractors. The result is a high-quality benchmark for evaluating multilingual and multimodal idiomatic language understanding.

The launch of \$Trump coin ignited a wave in meme coin investment. Copy trading, as a strategy-agnostic approach that eliminates the need for deep trading knowledge, quickly gains widespread popularity in the meme coin market. However, copy trading is not a guarantee of profitability due to the prevalence of manipulative bots, the uncertainty of the followed wallets' future performance, and the lag in trade execution. Recently, large language models (LLMs) have shown promise in financial applications by effectively understanding multi-modal data and producing explainable decisions. However, a single LLM struggles with complex, multi-faceted tasks such as asset allocation. These challenges are even more pronounced in cryptocurrency markets, where LLMs often lack sufficient domain-specific knowledge in their training data. To address these challenges, we propose an explainable multi-agent system for meme coin copy trading. Inspired by the structure of an asset management team, our system decomposes the complex task into subtasks and coordinates specialized agents to solve them collaboratively. Employing few-shot chain-of-though (CoT) prompting, each agent acquires professional meme coin trading knowledge, interprets multi-modal data, and generates explainable decisions. Using a dataset of 1,000 meme coin projects' transaction data, our empirical evaluation shows that the proposed multi-agent system outperforms both traditional machine learning models and single LLMs, achieving 73% and 70% precision in identifying high-quality meme coin projects and key opinion leader (KOL) wallets, respectively. The selected KOLs collectively generated a total profit of \$500,000 across these projects.

While recent research has systematically documented political orientation in large language models (LLMs), existing evaluations rely primarily on direct probing or demographic persona engineering to surface ideological biases. In social psychology, however, political ideology is also understood as a downstream consequence of fundamental moral intuitions. In this work, we investigate the causal relationship between moral values and political positioning by treating moral orientation as a controllable condition. Rather than simply assigning a demographic persona, we condition models to endorse or reject specific moral values and evaluate the resulting shifts on their political orientations, using the Political Compass Test. By treating moral values as lenses, we observe how moral conditioning actively steers model trajectories across economic and social dimensions. Our findings show that such conditioning induces pronounced, value-specific shifts in models' political coordinates. We further notice that these effects are systematically modulated by role framing and model scale, and are robust across alternative assessment instruments instantiating the same moral value. This highlights that effective alignment requires anchoring political assessments within the context of broader social values including morality, paving the way for more socially grounded alignment techniques.

Forecasting time series with extreme events is critical yet challenging due to their high variance, irregular dynamics, and sparse but high-impact nature. While existing methods excel in modeling dominant regular patterns, their performance degrades significantly during extreme events, constituting the primary source of forecasting errors in real-world applications. Although some approaches incorporate auxiliary signals to improve performance, they still fail to capture extreme events' complex temporal dynamics. To address these limitations, we propose M$^2$FMoE, an extreme-adaptive forecasting model that learns both regular and extreme patterns through multi-resolution and multi-view frequency modeling. It comprises three modules: (1) a multi-view frequency mixture-of-experts module assigns experts to distinct spectral bands in Fourier and Wavelet domains, with cross-view shared band splitter aligning frequency partitions and enabling inter-expert collaboration to capture both dominant and rare fluctuations; (2) a multi-resolution adaptive fusion module that hierarchically aggregates frequency features from coarse to fine resolutions, enhancing sensitivity to both short-term variations and sudden changes; (3) a temporal gating integration module that dynamically balances long-term trends and short-term frequency-aware features, improving adaptability to both regular and extreme temporal patterns. Experiments on real-world hydrological datasets with extreme patterns demonstrate that M$^2$FMoE outperforms state-of-the-art baselines without requiring extreme-event labels.

Data curation is a critical yet under-researched step in the machine translation training paradigm. To train translation systems, data acquisition relies primarily on human translations and digital parallel sources or, to a limited degree, synthetic generation. But, for low-resource languages, human translation to generate sufficient data is prohibitively expensive. Therefore, it is crucial to develop a framework that screens source sentences to form efficient parallel text, ensuring optimal MT system performance in low-resource environments. We approach this by evaluating English-Hindi bi-text to determine effective sentence selection strategies for optimal MT system training. Our extensively tested framework, (Lexical And Linguistically Informed Text Analysis) LALITA, targets source sentence selection using lexical and linguistic features to curate parallel corpora. We find that by training mostly on complex sentences from both existing and synthetic datasets, our method significantly improves translation quality. We test this by simulating low-resource data availabilty with curated datasets of 50K to 800K English sentences and report improved performances on all data sizes. LALITA demonstrates remarkable efficiency, reducing data needs by more than half across multiple languages (Hindi, Odia, Nepali, Norwegian Nynorsk, and German). This approach not only reduces MT systems training cost by reducing training data requirement, but also showcases LALITA's utility in data augmentation.

Large language models (LLMs) excel at semantic understanding, yet their ability to reconstruct internal structure from scrambled inputs remains underexplored. Sentence-level restoration is ill-posed for automated evaluation because multiple valid word orders often exist. We introduce OrderProbe, a deterministic benchmark for structural reconstruction using fixed four-character expressions in Chinese, Japanese, and Korean, which have a unique canonical order and thus support exact-match scoring. We further propose a diagnostic framework that evaluates models beyond recovery accuracy, including semantic fidelity, logical validity, consistency, robustness sensitivity, and information density. Experiments on twelve widely used LLMs show that structural reconstruction remains difficult even for frontier systems: zero-shot recovery frequently falls below 35%. We also observe a consistent dissociation between semantic recall and structural planning, suggesting that structural robustness is not an automatic byproduct of semantic competence.

Image generation models (IGMs), while capable of producing impressive and creative content, often memorize a wide range of undesirable concepts from their training data, leading to the reproduction of unsafe content such as NSFW imagery and copyrighted artistic styles. Such behaviors pose persistent safety and compliance risks in real-world deployments and cannot be reliably mitigated by post-hoc filtering, owing to the limited robustness of such mechanisms and a lack of fine-grained semantic control. Recent unlearning methods seek to erase harmful concepts at the model level, which exhibit the limitations of requiring costly retraining, degrading the quality of benign generations, or failing to withstand prompt paraphrasing and adversarial attacks. To address these challenges, we introduce SafeRedir, a lightweight inference-time framework for robust unlearning via prompt embedding redirection. Without modifying the underlying IGMs, SafeRedir adaptively routes unsafe prompts toward safe semantic regions through token-level interventions in the embedding space. The framework comprises two core components: a latent-aware multi-modal safety classifier for identifying unsafe generation trajectories, and a token-level delta generator for precise semantic redirection, equipped with auxiliary predictors for token masking and adaptive scaling to localize and regulate the intervention. Empirical results across multiple representative unlearning tasks demonstrate that SafeRedir achieves effective unlearning capability, high semantic and perceptual preservation, robust image quality, and enhanced resistance to adversarial attacks. Furthermore, SafeRedir generalizes effectively across a variety of diffusion backbones and existing unlearned models, validating its plug-and-play compatibility and broad applicability. Code and data are available at https://github.com/ryliu68/SafeRedir.

Recent advances in search-augmented large reasoning models (LRMs) enable the retrieval of external knowledge to reduce hallucinations in multistep reasoning. However, their ability to operate on graph-structured data, prevalent in domains such as e-commerce, social networks, and scientific citations, remains underexplored. Unlike plain text corpora, graphs encode rich topological signals that connect related entities and can serve as valuable priors for retrieval, enabling more targeted search and improved reasoning efficiency. Yet, effectively leveraging such structure poses unique challenges, including the difficulty of generating graph-expressive queries and ensuring reliable retrieval that balances structural and semantic relevance. To address this gap, we introduce GraphSearch, the first framework that extends search-augmented reasoning to graph learning, enabling zero-shot graph learning without task-specific fine-tuning. GraphSearch combines a Graph-aware Query Planner, which disentangles search space (e.g., 1-hop, multi-hop, or global neighbors) from semantic queries, with a Graph-aware Retriever, which constructs candidate sets based on topology and ranks them using a hybrid scoring function. We further instantiate two traversal modes: GraphSearch-R, which recursively expands neighborhoods hop by hop, and GraphSearch-F, which flexibly retrieves across local and global neighborhoods without hop constraints. Extensive experiments across diverse benchmarks show that GraphSearch achieves competitive or even superior performance compared to supervised graph learning methods, setting state-of-the-art results in zero-shot node classification and link prediction. These findings position GraphSearch as a flexible and generalizable paradigm for agentic reasoning over graphs.

Retrieval-Augmented Generation (RAG) pipelines must address challenges beyond simple single-document retrieval, such as interpreting visual elements (tables, charts, images), synthesizing information across documents, and providing accurate source grounding. Existing benchmarks fail to capture this complexity, often focusing on textual data, single-document comprehension, or evaluating retrieval and generation in isolation. We introduce ViDoRe v3, a comprehensive multimodal RAG benchmark featuring multi-type queries over visually rich document corpora. It covers 10 datasets across diverse professional domains, comprising ~26,000 document pages paired with 3,099 human-verified queries, each available in 6 languages. Through 12,000 hours of human annotation effort, we provide high-quality annotations for retrieval relevance, bounding box localization, and verified reference answers. Our evaluation of state-of-the-art RAG pipelines reveals that visual retrievers outperform textual ones, late-interaction models and textual reranking substantially improve performance, and hybrid or purely visual contexts enhance answer generation quality. However, current models still struggle with non-textual elements, open-ended queries, and fine-grained visual grounding. To encourage progress in addressing these challenges, the benchmark is released under a commercially permissive license at https://hf.co/vidore.

Multiple binary responses arise in many modern data-analytic problems. Although fitting separate logistic regressions for each response is computationally attractive, it ignores shared structure and can be statistically inefficient, especially in high-dimensional and class-imbalanced regimes. Low-rank models offer a natural way to encode latent dependence across tasks, but existing methods for binary data are largely likelihood-based and focus on pointwise classification rather than ranking performance. In this work, we propose a unified framework for learning with multiple binary responses that directly targets discrimination by minimizing a surrogate loss for the area under the ROC curve (AUC). The method aggregates pairwise AUC surrogate losses across responses while imposing a low-rank constraint on the coefficient matrix to exploit shared structure. We develop a scalable projected gradient descent algorithm based on truncated singular value decomposition. Exploiting the fact that the pairwise loss depends only on differences of linear predictors, we simplify computation and analysis. We establish non-asymptotic convergence guarantees, showing that under suitable regularity conditions, leading to linear convergence up to the minimax-optimal statistical precision. Extensive simulation studies demonstrate that the proposed method is robust in challenging settings such as label switching and data contamination and consistently outperforms likelihood-based approaches.

Heterogeneity within data distribution poses a challenge in many modern federated learning tasks. We formalize it as an optimization problem involving a computationally heavy composite under data similarity. By employing different sets of assumptions, we present several approaches to develop communication-efficient methods. An optimal algorithm is proposed for the convex case. The constructed theory is validated through a series of experiments across various problems.

The growing scale of online misinformation urgently demands Automated Fact-Checking (AFC). Existing benchmarks for evaluating AFC systems, however, are largely limited in terms of task scope, modalities, domain, language diversity, realism, or coverage of misinformation types. Critically, they are static, thus subject to data leakage as their claims enter the pretraining corpora of LLMs. As a result, benchmark performance no longer reliably reflects the actual ability to verify claims. We introduce Verified Theses and Statements (VeriTaS), the first dynamic benchmark for multimodal AFC, designed to remain robust under ongoing large-scale pretraining of foundation models. VeriTaS currently comprises 24,000 real-world claims from 108 professional fact-checking organizations across 54 languages, covering textual and audiovisual content. Claims are added quarterly via a fully automated seven-stage pipeline that normalizes claim formulation, retrieves original media, and maps heterogeneous expert verdicts to a novel, standardized, and disentangled scoring scheme with textual justifications. Through human evaluation, we demonstrate that the automated annotations closely match human judgments. We commit to update VeriTaS in the future, establishing a leakage-resistant benchmark, supporting meaningful AFC evaluation in the era of rapidly evolving foundation models. We will make the code and data publicly available.

Autonomous Driving Assistance Systems (ADAS) rely on extensive testing to ensure safety and reliability, yet road scenario datasets often contain redundant cases that slow down the testing process without improving fault detection. To address this issue, we present a novel test prioritization framework that reduces redundancy while preserving geometric and behavioral diversity. Road scenarios are clustered based on geometric and dynamic features of the ADAS driving behavior, from which representative cases are selected to guarantee coverage. Roads are finally prioritized based on geometric complexity, driving difficulty, and historical failures, ensuring that the most critical and challenging tests are executed first. We evaluate our framework on the OPENCAT dataset and the Udacity self-driving car simulator using two ADAS models. On average, our approach achieves an 89% reduction in test suite size while retaining an average of 79% of failed road scenarios. The prioritization strategy improves early failure detection by up to 95x compared to random baselines.

Experience intervention in web agents emerges as a promising technical paradigm, enhancing agent interaction capabilities by providing valuable insights from accumulated experiences. However, existing methods predominantly inject experience passively as global context before task execution, struggling to adapt to dynamically changing contextual observations during agent-environment interaction. We propose ExpSeek, which shifts experience toward step-level proactive seeking: (1) estimating step-level entropy thresholds to determine intervention timing using the model's intrinsic signals; (2) designing step-level tailor-designed experience content. Experiments on Qwen3-8B and 32B models across four challenging web agent benchmarks demonstrate that ExpSeek achieves absolute improvements of 9.3% and 7.5%, respectively. Our experiments validate the feasibility and advantages of entropy as a self-triggering signal, reveal that even a 4B small-scale experience model can significantly boost the performance of larger agent models.

For automated assessment of knee MRI scans, both accuracy and interpretability are essential for clinical use and adoption. Traditional radiomics rely on predefined features chosen at the population level; while more interpretable, they are often too restrictive to capture patient-specific variability and can underperform end-to-end deep learning (DL). To address this, we propose two complementary strategies that bring individuality and interpretability: radiomic fingerprints and healthy personas. First, a radiomic fingerprint is a dynamically constructed, patient-specific feature set derived from MRI. Instead of applying a uniform population-level signature, our model predicts feature relevance from a pool of candidate features and selects only those most predictive for each patient, while maintaining feature-level interpretability. This fingerprint can be viewed as a latent-variable model of feature usage, where an image-conditioned predictor estimates usage probabilities and a transparent logistic regression with global coefficients performs classification. Second, a healthy persona synthesises a pathology-free baseline for each patient using a diffusion model trained to reconstruct healthy knee MRIs. Comparing features extracted from pathological images against their personas highlights deviations from normal anatomy, enabling intuitive, case-specific explanations of disease manifestations. We systematically compare fingerprints, personas, and their combination across three clinical tasks. Experimental results show that both approaches yield performance comparable to or surpassing state-of-the-art DL models, while supporting interpretability at multiple levels. Case studies further illustrate how these perspectives facilitate human-explainable biomarker discovery and pathology localisation.

Vision modeling has advanced rapidly with Transformers, whose attention mechanisms capture visual dependencies but lack a principled account of how semantic information propagates spatially. We revisit this problem from a wave-based perspective: feature maps are treated as spatial signals whose evolution over an internal propagation time (aligned with network depth) is governed by an underdamped wave equation. In this formulation, spatial frequency-from low-frequency global layout to high-frequency edges and textures-is modeled explicitly, and its interaction with propagation time is controlled rather than implicitly fixed. We derive a closed-form, frequency-time decoupled solution and implement it as the Wave Propagation Operator (WPO), a lightweight module that models global interactions in O(N log N) time-far lower than attention. Building on WPO, we propose a family of WaveFormer models as drop-in replacements for standard ViTs and CNNs, achieving competitive accuracy across image classification, object detection, and semantic segmentation, while delivering up to 1.6x higher throughput and 30% fewer FLOPs than attention-based alternatives. Furthermore, our results demonstrate that wave propagation introduces a complementary modeling bias to heat-based methods, effectively capturing both global coherence and high-frequency details essential for rich visual semantics. Codes are available at: https://github.com/ZishanShu/WaveFormer.

This paper investigates symmetric composite binary quantum hypothesis testing (QHT), where the goal is to determine which of two uncertainty sets contains an unknown quantum state. While asymptotic error exponents for this problem are well-studied, the finite-sample regime remains poorly understood. We bridge this gap by characterizing the sample complexity -- the minimum number of state copies required to achieve a target error level. Specifically, we derive lower bounds that generalize the sample complexity of simple QHT and introduce new upper bounds for various uncertainty sets, including of both finite and infinite cardinalities. Notably, our upper and lower bounds match up to universal constants, providing a tight characterization of the sample complexity. Finally, we extend our analysis to the differentially private setting, establishing the sample complexity for privacy-preserving composite QHT.

We introduce the Ministral 3 series, a family of parameter-efficient dense language models designed for compute and memory constrained applications, available in three model sizes: 3B, 8B, and 14B parameters. For each model size, we release three variants: a pretrained base model for general-purpose use, an instruction finetuned, and a reasoning model for complex problem-solving. In addition, we present our recipe to derive the Ministral 3 models through Cascade Distillation, an iterative pruning and continued training with distillation technique. Each model comes with image understanding capabilities, all under the Apache 2.0 license.

Video is a powerful medium for communication and storytelling, yet reauthoring existing footage remains challenging. Even simple edits often demand expertise, time, and careful planning, constraining how creators envision and shape their narratives. Recent advances in generative AI suggest a new paradigm: what if editing a video were as straightforward as rewriting text? To investigate this, we present a tech probe and a study on text-driven video reauthoring. Our approach involves two technical contributions: (1) a generative reconstruction algorithm that reverse-engineers video into an editable text prompt, and (2) an interactive probe, Rewrite Kit, that allows creators to manipulate these prompts. A technical evaluation of the algorithm reveals a critical human-AI perceptual gap. A probe study with 12 creators surfaced novel use cases such as virtual reshooting, synthetic continuity, and aesthetic restyling. It also highlighted key tensions around coherence, control, and creative alignment in this new paradigm. Our work contributes empirical insights into the opportunities and challenges of text-driven video reauthoring, offering design implications for future co-creative video tools.

Sustainable water resource management in transboundary river basins is challenged by fragmented data, limited real-time access, and the complexity of integrating diverse information sources. This paper presents WaterCopilot-an AI-driven virtual assistant developed through collaboration between the International Water Management Institute (IWMI) and Microsoft Research for the Limpopo River Basin (LRB) to bridge these gaps through a unified, interactive platform. Built on Retrieval-Augmented Generation (RAG) and tool-calling architectures, WaterCopilot integrates static policy documents and real-time hydrological data via two custom plugins: the iwmi-doc-plugin, which enables semantic search over indexed documents using Azure AI Search, and the iwmi-api-plugin, which queries live databases to deliver dynamic insights such as environmental-flow alerts, rainfall trends, reservoir levels, water accounting, and irrigation data. The system features guided multilingual interactions (English, Portuguese, French), transparent source referencing, automated calculations, and visualization capabilities. Evaluated using the RAGAS framework, WaterCopilot achieves an overall score of 0.8043, with high answer relevancy (0.8571) and context precision (0.8009). Key innovations include automated threshold-based alerts, integration with the LRB Digital Twin, and a scalable deployment pipeline hosted on AWS. While limitations in processing non-English technical documents and API latency remain, WaterCopilot establishes a replicable AI-augmented framework for enhancing water governance in data-scarce, transboundary contexts. The study demonstrates the potential of this AI assistant to support informed, timely decision-making and strengthen water security in complex river basins.

Hallucinations in video-capable vision-language models (Video-VLMs) remain frequent and high-confidence, while existing uncertainty metrics often fail to align with correctness. We introduce VideoHEDGE, a modular framework for hallucination detection in video question answering that extends entropy-based reliability estimation from images to temporally structured inputs. Given a video-question pair, VideoHEDGE draws a baseline answer and multiple high-temperature generations from both clean clips and photometrically and spatiotemporally perturbed variants, then clusters the resulting textual outputs into semantic hypotheses using either Natural Language Inference (NLI)-based or embedding-based methods. Cluster-level probability masses yield three reliability scores: Semantic Entropy (SE), RadFlag, and Vision-Amplified Semantic Entropy (VASE). We evaluate VideoHEDGE on the SoccerChat benchmark using an LLM-as-a-judge to obtain binary hallucination labels. Across three 7B Video-VLMs (Qwen2-VL, Qwen2.5-VL, and a SoccerChat-finetuned model), VASE consistently achieves the highest ROC-AUC, especially at larger distortion budgets, while SE and RadFlag often operate near chance. We further show that embedding-based clustering matches NLI-based clustering in detection performance at substantially lower computational cost, and that domain fine-tuning reduces hallucination frequency but yields only modest improvements in calibration. The hedge-bench PyPI library enables reproducible and extensible benchmarking, with full code and experimental resources available at https://github.com/Simula/HEDGE#videohedge .

Intelligibility and accurate uncertainty estimation are crucial for reliable decision-making. In this paper, we propose EviNAM, an extension of evidential learning that integrates the interpretability of Neural Additive Models (NAMs) with principled uncertainty estimation. Unlike standard Bayesian neural networks and previous evidential methods, EviNAM enables, in a single pass, both the estimation of the aleatoric and epistemic uncertainty as well as explicit feature contributions. Experiments on synthetic and real data demonstrate that EviNAM matches state-of-the-art predictive performance. While we focus on regression, our method extends naturally to classification and generalized additive models, offering a path toward more intelligible and trustworthy predictions.

We introduce a two-stage multitask learning framework for analyzing Electroencephalography (EEG) signals that integrates denoising, dynamical modeling, and representation learning. In the first stage, a denoising autoencoder is trained to suppress artifacts and stabilize temporal dynamics, providing robust signal representations. In the second stage, a multitask architecture processes these denoised signals to achieve three objectives: motor imagery classification, chaotic versus non-chaotic regime discrimination using Lyapunov exponent-based labels, and self-supervised contrastive representation learning with NT-Xent loss. A convolutional backbone combined with a Transformer encoder captures spatial-temporal structure, while the dynamical task encourages sensitivity to nonlinear brain dynamics. This staged design mitigates interference between reconstruction and discriminative goals, improves stability across datasets, and supports reproducible training by clearly separating noise reduction from higher-level feature learning. Empirical studies show that our framework not only enhances robustness and generalization but also surpasses strong baselines and recent state-of-the-art methods in EEG decoding, highlighting the effectiveness of combining denoising, dynamical features, and self-supervised learning.

Convolutional neural networks are widely used in imaging and image recognition. Learning such networks from training data leads to the minimization of a non-convex function. This makes the analysis of standard optimization methods such as variants of (stochastic) gradient descent challenging. In this article we study the simplified setting of linear convolutional networks. We show that the gradient flow (to be interpreted as an abstraction of gradient descent) applied to the empirical risk defined via certain loss functions including the square loss always converges to a critical point, under a mild condition on the training data.

With the development of large language models (LLMs) in the field of programming, intelligent programming coaching systems have gained widespread attention. However, most research focuses on repairing the buggy code of programming learners without providing the underlying causes of the bugs. To address this gap, we introduce a novel task, namely \textbf{LPR} (\textbf{L}earner-Tailored \textbf{P}rogram \textbf{R}epair). We then propose a novel and effective framework, \textbf{\textsc{\MethodName{}}} (\textbf{L}earner-Tailored \textbf{S}olution \textbf{G}enerator), to enhance program repair while offering the bug descriptions for the buggy code. In the first stage, we utilize a repair solution retrieval framework to construct a solution retrieval database and then employ an edit-driven code retrieval approach to retrieve valuable solutions, guiding LLMs in identifying and fixing the bugs in buggy code. In the second stage, we propose a solution-guided program repair method, which fixes the code and provides explanations under the guidance of retrieval solutions. Moreover, we propose an Iterative Retrieval Enhancement method that utilizes evaluation results of the generated code to iteratively optimize the retrieval direction and explore more suitable repair strategies, improving performance in practical programming coaching scenarios. The experimental results show that our approach outperforms a set of baselines by a large margin, validating the effectiveness of our framework for the newly proposed LPR task.

The rapid growth of AI has fueled the expansion of accelerator- or GPU-based data centers. However, the rising operational energy consumption has emerged as a critical bottleneck and a major sustainability concern. Dynamic Voltage and Frequency Scaling (DVFS) is a well-known technique used to reduce energy consumption, and thus improve energy-efficiency, since it requires little effort and works with existing hardware. Reducing the energy consumption of training and inference of Large Language Models (LLMs) through DVFS or power capping is feasible: related work has shown energy savings can be significant, but at the cost of significant slowdowns. In this work, we focus on reducing waste in LLM operations: i.e., reducing energy consumption without losing performance. We propose a fine-grained, kernel-level, DVFS approach that explores new frequency configurations, and prove these save more energy than previous, pass- or iteration-level solutions. For example, for a GPT-3 training run, a pass-level approach could reduce energy consumption by 2% (without losing performance), while our kernel-level approach saves as much as 14.6% (with a 0.6% slowdown). We further investigate the effect of data and tensor parallelism, and show our discovered clock frequencies translate well for both. We conclude that kernel-level DVFS is a suitable technique to reduce waste in LLM operations, providing significant energy savings with negligible slow-down.

Deep Research Systems (DRS) aim to help users search the web, synthesize information, and deliver comprehensive investigative reports. However, how to rigorously evaluate these systems remains under-explored. Existing deep-research benchmarks often fall into two failure modes. Some do not adequately test a system's ability to analyze evidence and write coherent reports. Others rely on evaluation criteria that are either overly coarse or directly defined by LLMs (or both), leading to scores that can be biased relative to human experts and are hard to verify or interpret. To address these issues, we introduce Deep Research Bench II, a new benchmark for evaluating DRS-generated reports. It contains 132 grounded research tasks across 22 domains; for each task, a system must produce a long-form research report that is evaluated by a set of 9430 fine-grained binary rubrics in total, covering three dimensions: information recall, analysis, and presentation. All rubrics are derived from carefully selected expert-written investigative articles and are constructed through a four-stage LLM+human pipeline that combines automatic extraction with over 400 human-hours of expert review, ensuring that the criteria are atomic, verifiable, and aligned with human expert judgment. We evaluate several state-of-the-art deep-research systems on Deep Research Bench II and find that even the strongest models satisfy fewer than 50% of the rubrics, revealing a substantial gap between current DRSs and human experts.

Facade renovation offers a more sustainable alternative to full demolition, yet producing design proposals that preserve existing structures while expressing new intent remains challenging. Current workflows typically require detailed as-built modelling before design, which is time-consuming, labour-intensive, and often involves repeated revisions. To solve this issue, we propose a three-stage framework combining generative artificial intelligence (AI) and vision-language models (VLM) that directly processes rough structural sketch and textual descriptions to produce consistent renovation proposals. First, the input sketch is used by a fine-tuned VLM model to predict bounding boxes specifying where modifications are needed and which components should be added. Next, a stable diffusion model generates detailed sketches of new elements, which are merged with the original outline through a generative inpainting pipeline. Finally, ControlNet is employed to refine the result into a photorealistic image. Experiments on datasets and real industrial buildings indicate that the proposed framework can generate renovation proposals that preserve the original structure while improving facade detail quality. This approach effectively bypasses the need for detailed as-built modelling, enabling architects to rapidly explore design alternatives, iterate on early-stage concepts, and communicate renovation intentions with greater clarity.

We propose a novel method for sampling from unnormalized Boltzmann densities based on a probability-flow ordinary differential equation (ODE) derived from linear stochastic interpolants. The key innovation of our approach is the use of a sequence of Langevin samplers to enable efficient simulation of the flow. Specifically, these Langevin samplers are employed (i) to generate samples from the interpolant distribution at intermediate times and (ii) to construct, starting from these intermediate times, a robust estimator of the velocity field governing the flow ODE. For both applications of the Langevin diffusions, we establish convergence guarantees. Extensive numerical experiments demonstrate the efficiency of the proposed method on challenging multimodal distributions across a range of dimensions, as well as its effectiveness in Bayesian inference tasks.

Spike-Timing-Dependent Plasticity (STDP) provides a biologically grounded learning rule for spiking neural networks (SNNs), but its reliance on precise spike timing and pairwise updates limits fast learning of weights. We introduce a supervised extension of Spike Agreement-Dependent Plasticity (SADP), which replaces pairwise spike-timing comparisons with population-level agreement metrics such as Cohen's kappa. The proposed learning rule preserves strict synaptic locality, admits linear-time complexity, and enables efficient supervised learning without backpropagation, surrogate gradients, or teacher forcing. We integrate supervised SADP within hybrid CNN-SNN architectures, where convolutional encoders provide compact feature representations that are converted into Poisson spike trains for agreement-driven learning in the SNN. Extensive experiments on MNIST, Fashion-MNIST, CIFAR-10, and biomedical image classification tasks demonstrate competitive performance and fast convergence. Additional analyses show stable performance across broad hyperparameter ranges and compatibility with device-inspired synaptic update dynamics. Together, these results establish supervised SADP as a scalable, biologically grounded, and hardware-aligned learning paradigm for spiking neural networks.

Reinforcement Learning from Verifier Rewards (RLVR) has emerged as a widely used approach for post-training large language models on reasoning tasks, with group-based methods such as GRPO and its variants gaining broad adoption. These methods rely on group-relative advantage estimation to avoid learned critics, yet its theoretical properties remain poorly understood. In this work, we uncover a fundamental issue of group-based RL: the group-relative advantage estimator is inherently biased relative to the true (expected) advantage. We provide the first theoretical analysis showing that it systematically underestimates advantages for hard prompts and overestimates them for easy prompts, leading to imbalanced exploration and exploitation. To address this issue, we propose History-Aware Adaptive Difficulty Weighting (HA-DW), an adaptive reweighting scheme that adjusts advantage estimates based on an evolving difficulty anchor and training dynamics. Both theoretical analysis and experiments on five mathematical reasoning benchmarks demonstrate that HA-DW consistently improves performance when integrated into GRPO and its variants. Our results suggest that correcting biased advantage estimation is critical for robust and efficient RLVR training.

Few-shot class-incremental learning (FSCIL) receives significant attention from the public to perform classification continuously with a few training samples, which suffers from the key catastrophic forgetting problem. Existing methods usually employ an external memory to store previous knowledge and treat it with incremental classes equally, which cannot properly preserve previous essential knowledge. To solve this problem and inspired by recent distillation works on knowledge transfer, we propose a framework termed \textbf{C}onstrained \textbf{D}ataset \textbf{D}istillation (\textbf{CD$^2$}) to facilitate FSCIL, which includes a dataset distillation module (\textbf{DDM}) and a distillation constraint module~(\textbf{DCM}). Specifically, the DDM synthesizes highly condensed samples guided by the classifier, forcing the model to learn compacted essential class-related clues from a few incremental samples. The DCM introduces a designed loss to constrain the previously learned class distribution, which can preserve distilled knowledge more sufficiently. Extensive experiments on three public datasets show the superiority of our method against other state-of-the-art competitors.

The distinction between conditional, unconditional, and absolute convergence in infinite-dimensional spaces has fundamental implications for computational algorithms. While these concepts coincide in finite dimensions, the Dvoretzky-Rogers theorem establishes their strict separation in general Banach spaces. We present a comprehensive characterization theorem unifying seven equivalent conditions for unconditional convergence: permutation invariance, net convergence, subseries tests, sign stability, bounded multiplier properties, and weak uniform convergence. These theoretical results directly inform algorithmic stability analysis, governing permutation invariance in gradient accumulation for Stochastic Gradient Descent and justifying coefficient thresholding in frame-based signal processing. Our work bridges classical functional analysis with contemporary computational practice, providing rigorous foundations for order-independent and numerically robust summation processes.

Recent LLMs increasingly integrate reasoning mechanisms like Chain-of-Thought (CoT). However, this explicit reasoning exposes a new attack surface for inference-time backdoors, which inject malicious reasoning paths without altering model parameters. Because these attacks generate linguistically coherent paths, they effectively evade conventional detection. To address this, we propose STAR (State-Transition Amplification Ratio), a framework that detects backdoors by analyzing output probability shifts. STAR exploits the statistical discrepancy where a malicious input-induced path exhibits high posterior probability despite a low prior probability in the model's general knowledge. We quantify this state-transition amplification and employ the CUSUM algorithm to detect persistent anomalies. Experiments across diverse models (8B-70B) and five benchmark datasets demonstrate that STAR exhibits robust generalization capabilities, consistently achieving near-perfect performance (AUROC $\approx$ 1.0) with approximately $42\times$ greater efficiency than existing baselines. Furthermore, the framework proves robust against adaptive attacks attempting to bypass detection.

Movie screenplays are rich long-form narratives that interleave complex character relationships, temporally ordered events, and dialogue-driven interactions. While prior benchmarks target individual subtasks such as question answering or dialogue generation, they rarely evaluate whether models can construct a coherent story world and use it consistently across multiple forms of reasoning and generation. We introduce STAGE (Screenplay Text, Agents, Graphs and Evaluation), a unified benchmark for narrative understanding over full-length movie screenplays. STAGE defines four tasks: knowledge graph construction, scene-level event summarization, long-context screenplay question answering, and in-script character role-playing, all grounded in a shared narrative world representation. The benchmark provides cleaned scripts, curated knowledge graphs, and event- and character-centric annotations for 150 films across English and Chinese, enabling holistic evaluation of models' abilities to build world representations, abstract and verify narrative events, reason over long narratives, and generate character-consistent responses.

Time series forecasting is critical to real-world decision making, yet most existing approaches remain unimodal and rely on extrapolating historical patterns. While recent progress in large language models (LLMs) highlights the potential for multimodal forecasting, existing benchmarks largely provide retrospective or misaligned raw context, making it unclear whether such models meaningfully leverage textual inputs. In practice, human experts incorporate what-if scenarios with historical evidence, often producing distinct forecasts from the same observations under different scenarios. Inspired by this, we introduce What If TSF (WIT), a multimodal forecasting benchmark designed to evaluate whether models can condition their forecasts on contextual text, especially future scenarios. By providing expert-crafted plausible or counterfactual scenarios, WIT offers a rigorous testbed for scenario-guided multimodal forecasting. The benchmark is available at https://github.com/jinkwan1115/WhatIfTSF.

We introduce Temporal Fusion Nexus (TFN), a multi-modal and task-agnostic embedding model to integrate irregular time series and unstructured clinical narratives. We analysed TFN in post-kidney transplant (KTx) care, with a retrospective cohort of 3382 patients, on three key outcomes: graft loss, graft rejection, and mortality. Compared to state-of-the-art model in post KTx care, TFN achieved higher performance for graft loss (AUC 0.96 vs. 0.94) and graft rejection (AUC 0.84 vs. 0.74). In mortality prediction, TFN yielded an AUC of 0.86. TFN outperformed unimodal baselines (approx 10% AUC improvement over time series only baseline, approx 5% AUC improvement over time series with static patient data). Integrating clinical text improved performance across all tasks. Disentanglement metrics confirmed robust and interpretable latent factors in the embedding space, and SHAP-based attributions confirmed alignment with clinical reasoning. TFN has potential application in clinical tasks beyond KTx, where heterogeneous data sources, irregular longitudinal data, and rich narrative documentation are available.

Despite the recent advancements in NLP with the advent of Large Language Models (LLMs), Entity Linking (EL) for historical texts remains challenging due to linguistic variation, noisy inputs, and evolving semantic conventions. Existing solutions either require substantial training data or rely on domain-specific rules that limit scalability. In this paper, we present MHEL-LLaMo (Multilingual Historical Entity Linking with Large Language MOdels), an unsupervised ensemble approach combining a Small Language Model (SLM) and an LLM. MHEL-LLaMo leverages a multilingual bi-encoder (BELA) for candidate retrieval and an instruction-tuned LLM for NIL prediction and candidate selection via prompt chaining. Our system uses SLM's confidence scores to discriminate between easy and hard samples, applying an LLM only for hard cases. This strategy reduces computational costs while preventing hallucinations on straightforward cases. We evaluate MHEL-LLaMo on four established benchmarks in six European languages (English, Finnish, French, German, Italian and Swedish) from the 19th and 20th centuries. Results demonstrate that MHEL-LLaMo outperforms state-of-the-art models without requiring fine-tuning, offering a scalable solution for low-resource historical EL. The implementation of MHEL-LLaMo is available on Github.

Large models such as Vision Transformers (ViTs) have demonstrated remarkable superiority over smaller architectures like ResNet in few-shot classification, owing to their powerful representational capacity. However, fine-tuning such large models demands extensive GPU memory and prolonged training time, making them impractical for many real-world low-resource scenarios. To bridge this gap, we propose EfficientFSL, a query-only fine-tuning framework tailored specifically for few-shot classification with ViT, which achieves competitive performance while significantly reducing computational overhead. EfficientFSL fully leverages the knowledge embedded in the pre-trained model and its strong comprehension ability, achieving high classification accuracy with an extremely small number of tunable parameters. Specifically, we introduce a lightweight trainable Forward Block to synthesize task-specific queries that extract informative features from the intermediate representations of the pre-trained model in a query-only manner. We further propose a Combine Block to fuse multi-layer outputs, enhancing the depth and robustness of feature representations. Finally, a Support-Query Attention Block mitigates distribution shift by adjusting prototypes to align with the query set distribution. With minimal trainable parameters, EfficientFSL achieves state-of-the-art performance on four in-domain few-shot datasets and six cross-domain datasets, demonstrating its effectiveness in real-world applications.

Session-based recommendation systems must capture implicit user intents from sessions. However, existing models suffer from issues such as item interaction dominance and noisy sessions. We propose a multi-channel recommendation model, including a knowledge graph channel, a session hypergraph channel, and a session line graph channel, to capture information from multiple sources. Our model adaptively removes redundant edges in the knowledge graph channel to reduce noise. Knowledge graph representations cooperate with hypergraph representations for prediction to alleviate item dominance. We also generate in-session attention for denoising. Finally, we maximize mutual information between the hypergraph and line graph channels as an auxiliary task. Experiments demonstrate that our method enhances the accuracy of various recommendations, including e-commerce and multimedia recommendations. We release the code on GitHub for reproducibility.\footnote{https://github.com/hohehohe0509/DSR-HK}

Few-shot class-incremental learning (FSCIL) aims to continually adapt a model on a limited number of new-class examples, facing two well-known challenges: catastrophic forgetting and overfitting to new classes. Existing methods tend to freeze more parts of network components and finetune others with an extra memory during incremental sessions. These methods emphasize preserving prior knowledge to ensure proficiency in recognizing old classes, thereby mitigating catastrophic forgetting. Meanwhile, constraining fewer parameters can help in overcoming overfitting with the assistance of prior knowledge. Following previous methods, we retain more prior knowledge and propose a prior knowledge-infused neural network (PKI) to facilitate FSCIL. PKI consists of a backbone, an ensemble of projectors, a classifier, and an extra memory. In each incremental session, we build a new projector and add it to the ensemble. Subsequently, we finetune the new projector and the classifier jointly with other frozen network components, ensuring the rich prior knowledge is utilized effectively. By cascading projectors, PKI integrates prior knowledge accumulated from previous sessions and learns new knowledge flexibly, which helps to recognize old classes and efficiently learn new classes. Further, to reduce the resource consumption associated with keeping many projectors, we design two variants of the prior knowledge-infused neural network (PKIV-1 and PKIV-2) to trade off a balance between resource consumption and performance by reducing the number of projectors. Extensive experiments on three popular benchmarks demonstrate that our approach outperforms state-of-the-art methods.

Internet of underwater things (IoUT) is increasingly gathering attention with the aim of monitoring sea life and deep ocean environment, underwater surveillance as well as maintenance of underwater installments. However, conventional IoUT devices, reliant on battery power, face limitations in lifespan and pose environmental hazards upon disposal. This paper introduces a sustainable approach for simultaneous information uplink from the IoUT devices and acoustic energy transfer (AET) to the devices via an autonomous underwater vehicle (AUV), potentially enabling them to operate indefinitely. To tackle the time-sensitivity, we adopt age of information (AoI), and Jain's fairness index. We develop two deep-reinforcement learning (DRL) algorithms, offering a high-complexity, high-performance frequency division duplex (FDD) solution and a low-complexity, medium-performance time division duplex (TDD) approach. The results elucidate that the proposed FDD and TDD solutions significantly reduce the average AoI and boost the harvested energy as well as data collection fairness compared to baseline approaches.

We investigate a failure mode of large language models (LLMs) in which plain-text prompts elicit excessive outputs, a phenomenon we term Overflow. Unlike jailbreaks or prompt injection, Overflow arises under ordinary interaction settings and can lead to elevated serving cost, latency, and cross-user performance degradation, particularly when scaled across many requests. Beyond usability, the stakes are economic and environmental: unnecessary tokens increase per-request cost and energy consumption, compounding into substantial operational spend and carbon footprint at scale. Moreover, Overflow represents a practical vector for compute amplification and service degradation in shared environments. We introduce BenchOverflow, a model-agnostic benchmark of nine plain-text prompting strategies that amplify output volume without adversarial suffixes or policy circumvention. Using a standardized protocol with a fixed budget of 5000 new tokens, we evaluate nine open- and closed-source models and observe pronounced rightward shifts and heavy tails in length distributions. Cap-saturation rates (CSR@1k/3k/5k) and empirical cumulative distribution functions (ECDFs) quantify tail risk; within-prompt variance and cross-model correlations show that Overflow is broadly reproducible yet heterogeneous across families and attack vectors. A lightweight mitigation-a fixed conciseness reminder-attenuates right tails and lowers CSR for all strategies across the majority of models. Our findings position length control as a measurable reliability, cost, and sustainability concern rather than a stylistic quirk. By enabling standardized comparison of length-control robustness across models, BenchOverflow provides a practical basis for selecting deployments that minimize resource waste and operating expense, and for evaluating defenses that curb compute amplification without eroding task performance.

Safety-aligned language models systematically refuse harmful requests. While activation steering can modulate refusal, ablating the raw "refusal vector" calculated from contrastive harmful and harmless prompts often causes collateral damage and distribution drift. We argue this degradation occurs because the raw vector is polysemantic, entangling the refusal signal with core capability circuits and linguistic style. We introduce Surgical Refusal Ablation (SRA) to distill these steering directions. SRA constructs a registry of independent Concept Atoms representing protected capabilities and stylistic confounds, then uses ridge-regularized spectral residualization to orthogonalize the refusal vector against these directions. This yields a clean refusal direction that targets refusal-relevant structure while minimizing disruption to the model's semantic geometry. Across five models (Qwen3-VL and Ministral series), SRA achieves deep refusal reduction (0-2%) with negligible perplexity impact on Wikitext-2 (mean delta PPL approx. 0.02) and minimal distribution drift. Notably, standard ablation on Qwen3-VL-4B induces severe drift (first-token KL = 2.088), whereas SRA maintains the original distribution (KL = 0.044) while achieving the same 0% refusal rate. Using teacher-forced perplexity on GSM8K and MBPP as a high-resolution capability proxy, we show SRA preserves math and code distributions. These results suggest that common "model damage" is often "Ghost Noise," defined as the spectral bleeding of the dirty refusal direction into capability subspaces.

Map matching for sparse trajectories is a fundamental problem for many trajectory-based applications, e.g., traffic scheduling and traffic flow analysis. Existing methods for map matching are generally based on Hidden Markov Model (HMM) or encoder-decoder framework. However, these methods continue to face significant challenges when handling noisy or sparsely sampled GPS trajectories. To address these limitations, we propose DiffMM, an encoder-diffusion-based map matching framework that produces effective yet efficient matching results through a one-step diffusion process. We first introduce a road segment-aware trajectory encoder that jointly embeds the input trajectory and its surrounding candidate road segments into a shared latent space through an attention mechanism. Next, we propose a one step diffusion method to realize map matching through a shortcut model by leveraging the joint embedding of the trajectory and candidate road segments as conditioning context. We conduct extensive experiments on large-scale trajectory datasets, demonstrating that our approach consistently outperforms state-of-the-art map matching methods in terms of both accuracy and efficiency, particularly for sparse trajectories and complex road network topologies.

Conversational agents are increasingly used as support tools along mental therapeutic pathways with significant societal impacts. In particular, empathy is a key non-functional requirement in therapeutic contexts, yet current chatbot development practices provide no systematic means to specify or verify it. This paper envisions a framework integrating natural language processing and formal verification to deliver empathetic therapy chatbots. A Transformer-based model extracts dialogue features, which are then translated into a Stochastic Hybrid Automaton model of dyadic therapy sessions. Empathy-related properties can then be verified through Statistical Model Checking, while strategy synthesis provides guidance for shaping agent behavior. Preliminary results show that the formal model captures therapy dynamics with good fidelity and that ad-hoc strategies improve the probability of satisfying empathy requirements.

This paper incorporates the efficiency of automatic summarization and addresses the challenge of generating personalized summaries tailored to individual users' interests and requirements. To tackle this challenge, we introduce SummPilot, an interaction-based customizable summarization system. SummPilot leverages a large language model to facilitate both automatic and interactive summarization. Users can engage with the system to understand document content and personalize summaries through interactive components such as semantic graphs, entity clustering, and explainable evaluation. Our demo and user studies demonstrate SummPilot's adaptability and usefulness for customizable summarization.

Large language models frequently generate plausible but unfaithful summaries that users cannot verify against source text, a critical limitation in compliance-sensitive domains such as government and legal analysis. We present sui-1, a 24B parameter model that produces abstractive summaries with inline citations, enabling users to trace each claim to its source sentence. Our synthetic data pipeline combines chain-of-thought prompting with multi-stage verification, generating over 22,000 high-quality training examples across five languages from diverse sources including parliamentary documents, web text, and Wikipedia. Evaluation shows sui-1 significantly outperforms all tested open-weight baselines, including models with 3x more parameters. These results demonstrate that task-specific training substantially outperforms scale alone for citation-grounded summarization. Model weights and an interactive demo are publicly available.

Reinforcement Learning with Verifiable Rewards (RLVR) has become a standard paradigm for reasoning in Large Language Models. However, optimizing solely for final-answer correctness often drives models into aimless, verbose exploration, where they rely on exhaustive trial-and-error tactics rather than structured planning to reach solutions. While heuristic constraints like length penalties can reduce verbosity, they often truncate essential reasoning steps, creating a difficult trade-off between efficiency and verification. In this paper, we argue that discriminative capability is a prerequisite for efficient generation: by learning to distinguish valid solutions, a model can internalize a guidance signal that prunes the search space. We propose JudgeRLVR, a two-stage judge-then-generate paradigm. In the first stage, we train the model to judge solution responses with verifiable answers. In the second stage, we fine-tune the same model with vanilla generating RLVR initialized from the judge. Compared to Vanilla RLVR using the same math-domain training data, JudgeRLVR achieves a better quality--efficiency trade-off for Qwen3-30B-A3B: on in-domain math, it delivers about +3.7 points average accuracy gain with -42\% average generation length; on out-of-domain benchmarks, it delivers about +4.5 points average accuracy improvement, demonstrating enhanced generalization.

Distracted driving is a major cause of traffic collisions, calling for robust and scalable detection methods. Vision-language models (VLMs) enable strong zero-shot image classification, but existing VLM-based distracted driver detectors often underperform in real-world conditions. We identify subject-specific appearance variations (e.g., clothing, age, and gender) as a key bottleneck: VLMs entangle these factors with behavior cues, leading to decisions driven by who the driver is rather than what the driver is doing. To address this, we propose a subject decoupling framework that extracts a driver appearance embedding and removes its influence from the image embedding prior to zero-shot classification, thereby emphasizing distraction-relevant evidence. We further orthogonalize text embeddings via metric projection onto Stiefel manifold to improve separability while staying close to the original semantics. Experiments demonstrate consistent gains over prior baselines, indicating the promise of our approach for practical road-safety applications.

The conceptual design phase in architecture and urban planning, particularly building massing, is complex and heavily reliant on designer intuition and manual effort. To address this, we propose an automated framework for generating building massing based on functional requirements and site context. A primary obstacle to such data-driven methods has been the lack of suitable datasets. Consequently, we introduce the CoMa-20K dataset, a comprehensive collection that includes detailed massing geometries, associated economical and programmatic data, and visual representations of the development site within its existing urban context. We benchmark this dataset by formulating massing generation as a conditional task for Vision-Language Models (VLMs), evaluating both fine-tuned and large zero-shot models. Our experiments reveal the inherent complexity of the task while demonstrating the potential of VLMs to produce context-sensitive massing options. The dataset and analysis establish a foundational benchmark and highlight significant opportunities for future research in data-driven architectural design.

As the capabilities of large language model (LLM) agents continue to advance, their advanced social behaviors, such as cooperation, deception, and collusion, call for systematic evaluation. However, existing benchmarks often emphasize a single capability dimension or rely solely on behavioral outcomes, overlooking rich process information from agents' decision reasoning and communicative interactions. To address this gap, we propose M3-Bench, a multi-stage benchmark for mixed-motive games, together with a process-aware evaluation framework that conducts synergistic analysis across three modules: BTA (Behavioral Trajectory Analysis), RPA (Reasoning Process Analysis), and CCA (Communication Content Analysis). Furthermore, we integrate the Big Five personality model and Social Exchange Theory to aggregate multi-dimensional evidence into interpretable social behavior portraits, thereby characterizing agents' personality traits and capability profiles beyond simple task scores or outcome-based metrics. Experimental results show that M3-Bench can reliably distinguish diverse social behavior competencies across models, and it reveals that some models achieve seemingly reasonable behavioral outcomes while exhibiting pronounced inconsistencies in their reasoning and communication.

We provide a formal analytic proof for a class of non-canonical polynomial continued fractions representing π/4, originally conjectured by the Ramanujan Machine using algorithmic induction [4]. By establishing an explicit correspondence with the ratio of contiguous Gaussian hypergeometric functions 2F1(a, b; c; z), we show that these identities can be derived via a discrete sequence of equivalence transformations. We further prove that the conjectured integer coefficients constitute a symbolically minimal realization of the underlying analytic kernel. Stability analysis confirms that the resulting limit-periodic structures reside strictly within the Worpitzky convergence disk, ensuring absolute convergence. This work demonstrates that such algorithmically discovered identities are not isolated numerical artifacts, but are deeply rooted in the classical theory of hypergeometric transformations.

Digital platforms have an ever-expanding user base, and act as a hub for communication, business, and connectivity. However, this has also allowed for the spread of hate speech and misogyny. Artificial intelligence models have emerged as an effective solution for countering online hate speech but are under explored for low resource and code-mixed languages and suffer from a lack of interpretability. Explainable Artificial Intelligence (XAI) can enhance transparency in the decisions of deep learning models, which is crucial for a sensitive domain such as hate speech detection. In this paper, we present a multi-modal and explainable web application for detecting misogyny in text and memes in code-mixed Hindi and English. The system leverages state-of-the-art transformer-based models that support multilingual and multimodal settings. For text-based misogyny identification, the system utilizes XLM-RoBERTa (XLM-R) and multilingual Bidirectional Encoder Representations from Transformers (mBERT) on a dataset of approximately 4,193 comments. For multimodal misogyny identification from memes, the system utilizes mBERT + EfficientNet, and mBERT + ResNET trained on a dataset of approximately 4,218 memes. It also provides feature importance scores using explainability techniques including Shapley Additive Values (SHAP) and Local Interpretable Model Agnostic Explanations (LIME). The application aims to serve as a tool for both researchers and content moderators, to promote further research in the field, combat gender based digital violence, and ensure a safe digital space. The system has been evaluated using human evaluators who provided their responses on Chatbot Usability Questionnaire (CUQ) and User Experience Questionnaire (UEQ) to determine overall usability.

Autoregressive speech synthesis often adopts a left-to-right order, yet generation order is a modelling choice. We investigate decoding order through masked diffusion framework, which progressively unmasks positions and allows arbitrary decoding orders during training and inference. By interpolating between identity and random permutations, we show that randomness in decoding order affects speech quality. We further compare fixed strategies, such as \texttt{l2r} and \texttt{r2l} with adaptive ones, such as Top-$K$, finding that fixed-order decoding, including the dominating left-to-right approach, is suboptimal, while adaptive decoding yields better performance. Finally, since masked diffusion requires discrete inputs, we quantise acoustic representations and find that even 1-bit quantisation can support reasonably high-quality speech.

Few-shot class-incremental learning (FSCIL) aims to continuously recognize novel classes under limited data, which suffers from the key stability-plasticity dilemma: balancing the retention of old knowledge with the acquisition of new knowledge. To address this issue, we divide the task into two different stages and propose a framework termed Static-Dynamic Collaboration (SDC) to achieve a better trade-off between stability and plasticity. Specifically, our method divides the normal pipeline of FSCIL into Static Retaining Stage (SRS) and Dynamic Learning Stage (DLS), which harnesses old static and incremental dynamic class information, respectively. During SRS, we train an initial model with sufficient data in the base session and preserve the key part as static memory to retain fundamental old knowledge. During DLS, we introduce an extra dynamic projector jointly trained with the previous static memory. By employing both stages, our method achieves improved retention of old knowledge while continuously adapting to new classes. Extensive experiments on three public benchmarks and a real-world application dataset demonstrate that our method achieves state-of-the-art performance against other competitors.

Spike-timing-dependent plasticity (STDP) provides a biologically-plausible learning mechanism for spiking neural networks (SNNs); however, Hebbian weight updates in architectures with recurrent connections suffer from pathological weight dynamics: unbounded growth, catastrophic forgetting, and loss of representational diversity. We propose a neuromorphic regularization scheme inspired by the synaptic homeostasis hypothesis: periodic offline phases during which external inputs are suppressed, synaptic weights undergo stochastic decay toward a homeostatic baseline, and spontaneous activity enables memory consolidation. We demonstrate that this sleep-wake cycle prevents weight saturation while preserving learned structure. Empirically, we find that low to intermediate sleep durations (10-20\% of training) improve stability on MNIST-like benchmarks in our STDP-SNN model, without any data-specific hyperparameter tuning. In contrast, the same sleep intervention yields no measurable benefit for the surrogate-gradient spiking neural network (SG-SNN). Taken together, these results suggest that periodic, sleep-based renormalization may represent a fundamental mechanism for stabilizing local Hebbian learning in neuromorphic systems, while also indicating that special care is required when integrating such protocols with existing gradient-based optimization methods.

The development of reliable methods for multi-label classification (MLC) has become a prominent research direction in remote sensing (RS). As the scale of RS data continues to expand, annotation procedures increasingly rely on thematic products or crowdsourced procedures to reduce the cost of manual annotation. While cost-effective, these strategies often introduce multi-label noise in the form of partially incorrect annotations. In MLC, label noise arises as additive noise, subtractive noise, or a combination of both in the form of mixed noise. Previous work has largely overlooked this distinction and commonly treats noisy annotations as supervised signals, lacking mechanisms that explicitly adapt learning behavior to different noise types. To address this limitation, we propose NAR, a noise-adaptive regularization method that explicitly distinguishes between additive and subtractive noise within a semi-supervised learning framework. NAR employs a confidence-based label handling mechanism that dynamically retains label entries with high confidence, temporarily deactivates entries with moderate confidence, and corrects low confidence entries via flipping. This selective attenuation of supervision is integrated with early-learning regularization (ELR) to stabilize training and mitigate overfitting to corrupted labels. Experiments across additive, subtractive, and mixed noise scenarios demonstrate that NAR consistently improves robustness compared with existing methods. Performance improvements are most pronounced under subtractive and mixed noise, indicating that adaptive suppression and selective correction of noisy supervision provide an effective strategy for noise robust learning in RS MLC.

Table reasoning, a task to answer questions by reasoning over data presented in tables, is an important topic due to the prevalence of knowledge stored in tabular formats. Recent solutions use Large Language Models (LLMs), exploiting the semantic understanding and reasoning capabilities of LLMs. A common paradigm of such solutions linearizes tables to form plain texts that are served as input to LLMs. This paradigm has critical issues. It loses table structures, lacks explicit reasoning paths for result explainability, and is subject to the "lost-in-the-middle" issue. To address these issues, we propose Table Graph Reasoner (TABGR), a training-free model that represents tables as an Attributed Table Graph (ATG). The ATG explicitly preserves row-column-cell structures while enabling graph-based reasoning for explainability. We further propose a Question-Guided Personalized PageRank (QG-PPR) mechanism to rerank tabular data and mitigate the lost-in-the-middle issue. Extensive experiments on two commonly used benchmarks show that TABGR consistently outperforms state-of-the-art models by up to 9.7% in accuracy. Our code will be made publicly available upon publication.

Steering Large Language Models (LLMs) through activation interventions has emerged as a lightweight alternative to fine-tuning for alignment and personalization. Recent work on Bi-directional Preference Optimization (BiPO) shows that dense steering vectors can be learned directly from preference data in a Direct Preference Optimization (DPO) fashion, enabling control over truthfulness, hallucinations, and safety behaviors. However, dense steering vectors often entangle multiple latent factors due to neuron multi-semanticity, limiting their effectiveness and stability in fine-grained settings such as cultural alignment, where closely related values and behaviors (e.g., among Middle Eastern cultures) must be distinguished. In this paper, we propose Yet another Policy Optimization (YaPO), a \textit{reference-free} method that learns \textit{sparse steering vectors} in the latent space of a Sparse Autoencoder (SAE). By optimizing sparse codes, YaPO produces disentangled, interpretable, and efficient steering directions. Empirically, we show that YaPO converges faster, achieves stronger performance, and exhibits improved training stability compared to dense steering baselines. Beyond cultural alignment, YaPO generalizes to a range of alignment-related behaviors, including hallucination, wealth-seeking, jailbreak, and power-seeking. Importantly, YaPO preserves general knowledge, with no measurable degradation on MMLU. Overall, our results show that YaPO provides a general recipe for efficient, stable, and fine-grained alignment of LLMs, with broad applications to controllability and domain adaptation. The associated code and data are publicly available\footnote{https://github.com/MBZUAI-Paris/YaPO}.

Effective memory management is essential for large language model agents to navigate long-horizon tasks. Recent research has explored using Reinforcement Learning to develop specialized memory manager agents. However, existing approaches rely on final task performance as the primary reward, which results in severe reward sparsity and ineffective credit assignment, providing insufficient guidance for individual memory operations. To this end, we propose Fine-Mem, a unified framework designed for fine-grained feedback alignment. First, we introduce a Chunk-level Step Reward to provide immediate step-level supervision via auxiliary chunk-specific question answering tasks. Second, we devise Evidence-Anchored Reward Attribution to redistribute global rewards by anchoring credit to key memory operations, based on the specific memory items utilized as evidence in reasoning. Together, these components enable stable policy optimization and align local memory operations with the long-term utility of memory. Experiments on Memalpha and MemoryAgentBench demonstrate that Fine-Mem consistently outperforms strong baselines, achieving superior success rates across various sub-tasks. Further analysis reveals its adaptability and strong generalization capabilities across diverse model configurations and backbones.

The advent of Large Multimodal Models (LMMs) offers a promising technology to tackle the limitations of modular design in autonomous driving, which often falters in open-world scenarios requiring sustained environmental understanding and logical reasoning. Besides, embodied artificial intelligence facilitates policy optimization through closed-loop interactions to achieve the continuous learning capability, thereby advancing autonomous driving toward embodied intelligent (El) driving. However, such capability will be constrained by relying solely on LMMs to enhance EI driving without joint decision-making. This article introduces a novel semantics and policy dual-driven hybrid decision framework to tackle this challenge, ensuring continuous learning and joint decision. The framework merges LMMs for semantic understanding and cognitive representation, and deep reinforcement learning (DRL) for real-time policy optimization. We starts by introducing the foundational principles of EI driving and LMMs. Moreover, we examine the emerging opportunities this framework enables, encompassing potential benefits and representative use cases. A case study is conducted experimentally to validate the performance superiority of our framework in completing lane-change planning task. Finally, several future research directions to empower EI driving are identified to guide subsequent work.

Reinforcement Learning with Verifiable Rewards (RLVR) has driven substantial progress in reasoning-intensive domains like mathematics. However, optimizing open-ended generation remains challenging due to the lack of ground truth. While rubric-based evaluation offers a structured proxy for verification, existing methods suffer from scalability bottlenecks and coarse criteria, resulting in a supervision ceiling effect. To address this, we propose an automated Coarse-to-Fine Rubric Generation framework. By synergizing principle-guided synthesis, multi-model aggregation, and difficulty evolution, our approach produces comprehensive and highly discriminative criteria capable of capturing the subtle nuances. Based on this framework, we introduce RubricHub, a large-scale ($\sim$110k) and multi-domain dataset. We validate its utility through a two-stage post-training pipeline comprising Rubric-based Rejection Sampling Fine-Tuning (RuFT) and Reinforcement Learning (RuRL). Experimental results demonstrate that RubricHub unlocks significant performance gains: our post-trained Qwen3-14B achieves state-of-the-art (SOTA) results on HealthBench (69.3), surpassing proprietary frontier models such as GPT-5. The code and data will be released soon.

Group Relative Policy Optimization (GRPO) significantly enhances the reasoning performance of Large Language Models (LLMs). However, this success heavily relies on expensive external verifiers or human rules. Such dependency not only leads to significant computational costs and training latency, but also yields sparse rewards that hinder optimization efficiency. To address these challenges, we propose Latent-GRPO, a framework that derives intrinsic rewards directly from latent space geometry. Crucially, our empirical analysis reveals a compelling geometric property: terminal token representations of correct reasoning trajectories form dense clusters with high intra-class similarity, whereas incorrect trajectories remain scattered as outliers. In light of this discovery, we introduce the Iterative Robust Centroid Estimation (IRCE) algorithm, which generates dense, continuous rewards by mitigating magnitude fluctuations via spherical projection and estimating a robust ``truth centroid'' through iterative aggregation. Experimental results on multiple datasets show that our method maintains model performance while achieving a training speedup of over 2x compared to baselines. Furthermore, extensive results demonstrate strong generalization ability and robustness. The code will be released soon.

Online on-policy preference learning algorithms for language model alignment such as online direct policy optimization (DPO) can significantly outperform their offline counterparts. We provide a theoretical explanation for this phenomenon by analyzing how the sampling policy's coverage evolves throughout on-policy training. We propose and rigorously justify the \emph{coverage improvement principle}: with sufficient batch size, each update moves into a region around the target where coverage is uniformly better, making subsequent data increasingly informative and enabling rapid convergence. In the contextual bandit setting with Bradley-Terry preferences and linear softmax policy class, we show that on-policy DPO converges exponentially in the number of iterations for batch size exceeding a generalized coverage threshold. In contrast, any learner restricted to offline samples from the initial policy suffers a slower minimax rate, leading to a sharp separation in total sample complexity. Motivated by this analysis, we further propose a simple hybrid sampler based on a novel \emph{preferential} G-optimal design, which removes dependence on coverage and guarantees convergence in just two rounds. Finally, we develop principled on-policy schemes for reward distillation in the general function class setting, and show faster noiseless rates under an alternative deviation-based notion of coverage. Experimentally, we confirm that on-policy DPO and our proposed reward distillation algorithms outperform their off-policy counterparts and enjoy stable, monotonic performance gains across iterations.

Tax code prediction is a crucial yet underexplored task in automating invoicing and compliance management for large-scale e-commerce platforms. Each product must be accurately mapped to a node within a multi-level taxonomic hierarchy defined by national standards, where errors lead to financial inconsistencies and regulatory risks. This paper presents Taxon, a semantically aligned and expert-guided framework for hierarchical tax code prediction. Taxon integrates (i) a feature-gating mixture-of-experts architecture that adaptively routes multi-modal features across taxonomy levels, and (ii) a semantic consistency model distilled from large language models acting as domain experts to verify alignment between product titles and official tax definitions. To address noisy supervision in real business records, we design a multi-source training pipeline that combines curated tax databases, invoice validation logs, and merchant registration data to provide both structural and semantic supervision. Extensive experiments on the proprietary TaxCode dataset and public benchmarks demonstrate that Taxon achieves state-of-the-art performance, outperforming strong baselines. Further, an additional full hierarchical paths reconstruction procedure significantly improves structural consistency, yielding the highest overall F1 scores. Taxon has been deployed in production within Alibaba's tax service system, handling an average of over 500,000 tax code queries per day and reaching peak volumes above five million requests during business event with improved accuracy, interpretability, and robustness.

Large Language Models (LLMs) are increasingly integrated into academic research pipelines; however, the Terms of Service governing their use remain under-examined. We present a comparative analysis of the Terms of Service of five major LLM providers (Anthropic, DeepSeek, Google, OpenAI, and xAI) collected in November 2025. Our analysis reveals substantial variation in the stringency and specificity of usage restrictions for general users and researchers. We identify specific complexities for researchers in security research, computational social sciences, and psychological studies. We identify `regulatory gray areas' where Terms of Service create uncertainty for legitimate use. We contribute a publicly available resource comparing terms across platforms (OSF) and discuss implications for general users and researchers navigating this evolving landscape.

With large language models demonstrating significant potential in code generation tasks, their application to onboard control of resource-constrained Unmanned Aerial Vehicles has emerged as an important research direction. However, a notable contradiction exists between the high resource consumption of large models and the real-time, lightweight requirements of UAV platforms. This paper proposes an integrated approach that combines knowledge distillation, chain-of-thought guidance, and supervised fine-tuning for UAV multi-SDK control tasks, aiming to efficiently transfer complex reasoning and code generation capabilities to smaller models. Firstly, a high-quality dataset covering various mainstream UAV SDKs is constructed, featuring instruction-code-reasoning chains, and incorporates counterfactual negative samples for data augmentation, guiding the model to learn the end-to-end logic from instruction parsing to code generation. Secondly, leveraging DeepSeek-Coder-V2-Lite quantized via QLoRA as the teacher model, and based on a hybrid black-box and white-box distillation strategy, high-quality chain-of-thought soft labels are generated. These are combined with a weighted cross-entropy loss using hard labels to transfer complex reasoning capabilities to the smaller student model. Finally, through prompt tuning engineering optimized for the UAV control scenario, the model performance on core tasks such as SDK type recognition and function call matching is enhanced. Experimental results indicate that the distilled lightweight model maintains high code generation accuracy while achieving significant improvements in deployment and inference efficiency, effectively demonstrating the feasibility and superiority of our approach in achieving precise and lightweight intelligent control for UAVs

Web Agents are increasingly deployed to perform complex tasks in real web environments, yet their security evaluation remains fragmented and difficult to standardize. We present WebTrap Park, an automated platform for systematic security evaluation of Web Agents through direct observation of their concrete interactions with live web pages. WebTrap Park instantiates three major sources of security risk into 1,226 executable evaluation tasks and enables action based assessment without requiring agent modification. Our results reveal clear security differences across agent frameworks, highlighting the importance of agent architecture beyond the underlying model. WebTrap Park is publicly accessible at https://security.fudan.edu.cn/webagent and provides a scalable foundation for reproducible Web Agent security evaluation.

Partial differential equations (PDEs) are a central tool for modeling the dynamics of physical, engineering, and materials systems, but high-fidelity simulations are often computationally expensive. At the same time, many scientific applications can be viewed as the evolution of spatially distributed fields, making data-driven forecasting of such fields a core task in scientific machine learning. In this work we study autoregressive deep-learning surrogates for two-dimensional PDE dynamics on periodic domains, focusing on generalization to out-of-distribution initial conditions within a fixed PDE and parameter regime and on strict small-data settings with at most $\mathcal{O}(10^2)$ simulated trajectories per system. We introduce a multi-channel U-Net [...], evaluate it on five qualitatively different PDE families and compare it to ViT, AFNO, PDE-Transformer, and KAN-UNet under a common training setup. Across all datasets, me-UNet matches or outperforms these more complex architectures in terms of field-space error, spectral similarity, and physics-based metrics for in-distribution rollouts, while requiring substantially less training time. It also generalizes qualitatively to unseen initial conditions with as few as $\approx 20$ training simulations. A data-efficiency study and Grad-CAM analysis further suggest that, in small-data periodic 2D PDE settings, convolutional architectures with inductive biases aligned to locality and periodic boundary conditions remain strong contenders for accurate and moderately out-of-distribution-robust surrogate modeling.

Large language models are increasingly trained via reinforcement learning for personalized recommendation tasks, but standard methods like GRPO rely on sparse, sequence-level rewards that create a credit assignment gap, obscuring which tokens drive success. This gap is especially problematic when models must infer latent user intent from under-specified language without ground truth labels, a reasoning pattern rarely seen during pretraining. We introduce Owen-Shapley Policy Optimization (OSPO), a framework that redistributes sequence-level advantages based on tokens' marginal contributions to outcomes. Unlike value-model-based methods requiring additional computation, OSPO employs potential-based reward shaping via Shapley-Owen attributions to assign segment-level credit while preserving the optimal policy, learning directly from task feedback without parametric value models. By forming coalitions of semantically coherent units (phrases describing product attributes or sentences capturing preferences), OSPO identifies which response parts drive performance. Experiments on Amazon ESCI and H&M Fashion datasets show consistent gains over baselines, with notable test-time robustness to out-of-distribution retrievers unseen during training.

Recent advances in large language models (LLMs) demonstrate their potential as educational tutors. However, different tutoring strategies benefit different student personalities, and mismatches can be counterproductive to student outcomes. Despite this, current LLM tutoring systems do not take into account student personality traits. To address this problem, we first construct a taxonomy that links pedagogical methods to personality profiles, based on pedagogical literature. We simulate student-teacher conversations and use our framework to let the LLM tutor adjust its strategy to the simulated student personality. We evaluate the scenario with human teachers and find that they consistently prefer our approach over two baselines. Our method also increases the use of less common, high-impact strategies such as role-playing, which human and LLM annotators prefer significantly. Our findings pave the way for developing more personalized and effective LLM use in educational applications.

Objectives: To overcome challenges in diagnosing pericoronitis on panoramic radiographs, an AI-assisted assessment system integrating anatomical localization, pathological classification, and interpretability. Methods: A two-stage deep learning pipeline was implemented. The first stage used YOLOv8 to detect third molars and classify their anatomical positions and angulations based on Winter's classification. Detected regions were then fed into a second-stage classifier, a modified ResNet-50 architecture, for detecting radiographic features suggestive of pericoronitis. To enhance clinical trust, Grad-CAM was used to highlight key diagnostic regions on the radiographs. Results: The YOLOv8 component achieved 92% precision and 92.5% mean average precision. The ResNet-50 classifier yielded F1-scores of 88% for normal cases and 86% for pericoronitis. Radiologists reported 84% alignment between Grad-CAM and their diagnostic impressions, supporting the radiographic relevance of the interpretability output. Conclusion: The system shows strong potential for AI-assisted panoramic assessment, with explainable AI features that support clinical confidence.

Scaling large models requires optimization strategies that ensure rapid convergence grounded in stability. Maximal Update Parametrization ($\boldsymbolμ$P) provides a theoretical safeguard for width-invariant $Θ(1)$ activation control, whereas emerging optimizers like Muon are only ``half-aligned'' with these constraints: they control updates but allow weights to drift. To address this limitation, we introduce the \textbf{Spectral Sphere Optimizer (SSO)}, which enforces strict module-wise spectral constraints on both weights and their updates. By deriving the steepest descent direction on the spectral sphere, SSO realizes a fully $\boldsymbolμ$P-aligned optimization process. To enable large-scale training, we implement SSO as an efficient parallel algorithm within Megatron. Through extensive pretraining on diverse architectures, including Dense 1.7B, MoE 8B-A1B, and 200-layer DeepNet models, SSO consistently outperforms AdamW and Muon. Furthermore, we observe significant practical stability benefits, including improved MoE router load balancing, suppressed outliers, and strictly bounded activations.

Creativity in artificial intelligence is most often addressed through evaluative frameworks that aim to measure novelty, diversity, or usefulness in generated outputs. While such approaches have provided valuable insights into the behavior of modern generative models, they largely treat creativity as a property to be assessed rather than as a phenomenon to be explicitly modeled. In parallel, recent advances in large-scale generative systems, particularly multimodal architectures, have demonstrated increasingly sophisticated forms of pattern recombination, raising questions about the nature and limits of machine creativity. This paper proposes a generative perspective on creativity in AI, framing it as an emergent property of domain-limited generative models embedded within bounded informational environments. Rather than introducing new evaluative criteria, we focus on the structural and contextual conditions under which creative behaviors arise. We introduce a conceptual decomposition of creativity into four interacting components-pattern-based generation, induced world models, contextual grounding, and arbitrarity, and examine how these components manifest in multimodal generative systems. By grounding creativity in the interaction between generative dynamics and domain-specific representations, this work aims to provide a technical framework for studying creativity as an emergent phenomenon in AI systems, rather than as a post hoc evaluative label.

Mixture-of-Experts (MoE) architectures decouple model capacity from per-token computation, enabling scaling beyond the computational limits imposed by dense scaling laws. Yet how MoE architectures shape knowledge acquisition during pre-training, and how this process differs from dense architectures, remains unknown. To address this issue, we introduce Gated-LPI (Log-Probability Increase), a neuron-level attribution metric that decomposes log-probability increase across neurons. We present a time-resolved comparison of knowledge acquisition dynamics in MoE and dense architectures, tracking checkpoints over 1.2M training steps (~ 5.0T tokens) and 600K training steps (~ 2.5T tokens), respectively. Our experiments uncover three patterns: (1) Low-entropy backbone. The top approximately 1% of MoE neurons capture over 45% of positive updates, forming a high-utility core, which is absent in the dense baseline. (2) Early consolidation. The MoE model locks into a stable importance profile within < 100K steps, whereas the dense model remains volatile throughout training. (3) Functional robustness. Masking the ten most important MoE attention heads reduces relational HIT@10 by < 10%, compared with > 50% for the dense model, showing that sparsity fosters distributed -- rather than brittle -- knowledge storage. These patterns collectively demonstrate that sparsity fosters an intrinsically stable and distributed computational backbone from early in training, helping bridge the gap between sparse architectures and training-time interpretability.

This paper presents a Qualitative model for Reasoning about Object Rotations (QOR) which is applied to solve the Cube Comparison Test (CCT) by Ekstrom et al. (1976). A conceptual neighborhood graph relating the Rotation movement to the Location change and the Orientation change (CNGRLO) of the features on the cube sides has been built and it produces composition tables to calculate inferences for reasoning about rotations.

TWG 5 focused on developing and implementing effective strategies for enhancing AI literacy and agency of teachers, equipping them with the knowledge and skills necessary to integrate AI into their teaching practices. Explorations covered curriculum design, professional development programs, practical classroom applications, and policy guidelines aiming to empower educators to confidently utilize AI tools and foster a deeper understanding of AI concepts among students.

Pre-trained diffusion models have emerged as powerful generative priors for both unconditional and conditional sample generation, yet their outputs often deviate from the characteristics of user-specific target data. Such mismatches are especially problematic in domain adaptation tasks, where only a few reference examples are available and retraining the diffusion model is infeasible. Existing inference-time guidance methods can adjust sampling trajectories, but they typically optimize surrogate objectives such as classifier likelihoods rather than directly aligning with the target distribution. We propose MMD Guidance, a training-free mechanism that augments the reverse diffusion process with gradients of the Maximum Mean Discrepancy (MMD) between generated samples and a reference dataset. MMD provides reliable distributional estimates from limited data, exhibits low variance in practice, and is efficiently differentiable, which makes it particularly well-suited for the guidance task. Our framework naturally extends to prompt-aware adaptation in conditional generation models via product kernels. Also, it can be applied with computational efficiency in latent diffusion models (LDMs), since guidance is applied in the latent space of the LDM. Experiments on synthetic and real-world benchmarks demonstrate that MMD Guidance can achieve distributional alignment while preserving sample fidelity.

In high-order finite element analysis for elasticity, matrix-free (PA) methods are a key technology for overcoming the memory bottleneck of traditional Full Assembly (FA). However, existing implementations fail to fully exploit the special structure of modern CPU architectures and tensor-product elements, causing their performance "sweet spot" to anomalously remain at the low order of $p \approx 2$, which severely limits the potential of high-order methods. To address this challenge, we design and implement a highly optimized PA operator within the MFEM framework, deeply integrated with a Geometric Multigrid (GMG) preconditioner. Our multi-level optimization strategy includes replacing the original $O(p^6)$ generic algorithm with an efficient $O(p^4)$ one based on tensor factorization, exploiting Voigt symmetry to reduce redundant computations for the elasticity problem, and employing macro-kernel fusion to enhance data locality and break the memory bandwidth bottleneck. Extensive experiments on mainstream x86 and ARM architectures demonstrate that our method successfully shifts the performance "sweet spot" to the higher-order region of $p \ge 6$. Compared to the MFEM baseline, the optimized core operator (kernel) achieves speedups of 7x to 83x, which translates to a 3.6x to 16.8x end-to-end performance improvement in the complete solution process. This paper provides a validated and efficient practical path for conducting large-scale, high-order elasticity simulations on mainstream CPU hardware.

We introduce Geo-NVS-w, a geometry-aware framework for high-fidelity novel view synthesis from unstructured, in-the-wild image collections. While existing in-the-wild methods already excel at novel view synthesis, they often lack geometric grounding on complex surfaces, sometimes producing results that contain inconsistencies. Geo-NVS-w addresses this limitation by leveraging an underlying geometric representation based on a Signed Distance Function (SDF) to guide the rendering process. This is complemented by a novel Geometry-Preservation Loss which ensures that fine structural details are preserved. Our framework achieves competitive rendering performance, while demonstrating a 4-5x reduction reduction in energy consumption compared to similar methods. We demonstrate that Geo-NVS-w is a robust method for in-the-wild NVS, yielding photorealistic results with sharp, geometrically coherent details.

In quantum software engineering (QSE), quantum software testing (QST) has attracted increasing attention as quantum software systems grow in scale and complexity. Since QST evaluates quantum programs through execution under designed test inputs, empirical studies are widely used to assess the effectiveness of testing approaches. However, the design and reporting of empirical studies in QST remain highly diverse, and a shared methodological understanding has yet to emerge, making it difficult to interpret results and compare findings across studies. This paper presents a methodological analysis of empirical studies in QST through a systematic examination of 59 primary studies identified from a literature pool of size 384. We organize our analysis around ten research questions that cover key methodological dimensions of QST empirical studies, including objects under test, baseline comparison, testing setup, experimental configuration, and tool and artifact support. Through cross-study analysis along these dimensions, we characterize current empirical practices in QST, identify recurring limitations and inconsistencies, and highlight open methodological challenges. Based on our findings, we derive insights and recommendations to inform the design, execution, and reporting of future empirical studies in QST.

While dense retrieval models have achieved remarkable success, rigorous evaluation of their sensitivity to the position of relevant information (i.e., position bias) remains largely unexplored. Existing benchmarks typically employ position-agnostic relevance labels, conflating the challenge of processing long contexts with the bias against specific evidence locations. To address this challenge, we introduce PosIR (Position-Aware Information Retrieval), a comprehensive benchmark designed to diagnose position bias in diverse retrieval scenarios. PosIR comprises 310 datasets spanning 10 languages and 31 domains, constructed through a rigorous pipeline that ties relevance to precise reference spans, enabling the strict disentanglement of document length from information position. Extensive experiments with 10 state-of-the-art embedding models reveal that: (1) Performance on PosIR in long-context settings correlates poorly with the MMTEB benchmark, exposing limitations in current short-text benchmarks; (2) Position bias is pervasive and intensifies with document length, with most models exhibiting primacy bias while certain models show unexpected recency bias; (3) Gradient-based saliency analysis further uncovers the distinct internal attention mechanisms driving these positional preferences. In summary, PosIR serves as a valuable diagnostic framework to foster the development of position-robust retrieval systems.

Sequential recommender systems must model long-range user behavior while operating under strict memory and latency constraints. Transformer-based approaches achieve strong accuracy but suffer from quadratic attention complexity, forcing aggressive truncation of user histories and limiting their practicality for long-horizon modeling. This paper presents HoloMambaRec, a lightweight sequential recommendation architecture that combines holographic reduced representations for attribute-aware embedding with a selective state space encoder for linear-time sequence processing. Item and attribute information are bound using circular convolution, preserving embedding dimensionality while encoding structured metadata. A shallow selective state space backbone, inspired by recent Mamba-style models, enables efficient training and constant-time recurrent inference. Experiments on Amazon Beauty and MovieLens-1M datasets demonstrate that HoloMambaRec consistently outperforms SASRec and achieves competitive performance with GRU4Rec under a constrained 10-epoch training budget, while maintaining substantially lower memory complexity. The design further incorporates forward-compatible mechanisms for temporal bundling and inference-time compression, positioning HoloMambaRec as a practical and extensible alternative for scalable, metadata-aware sequential recommendation.

Increasing levels of anthropogenic noise from ships contribute significantly to underwater sound pollution, posing risks to marine ecosystems. This makes monitoring crucial to understand and quantify the impact of the ship radiated noise. Passive Acoustic Monitoring (PAM) systems are widely deployed for this purpose, generating years of underwater recordings across diverse soundscapes. Manual analysis of such large-scale data is impractical, motivating the need for automated approaches based on machine learning. Recent advances in automatic Underwater Acoustic Target Recognition (UATR) have largely relied on supervised learning, which is constrained by the scarcity of labeled data. Transfer Learning (TL) offers a promising alternative to mitigate this limitation. In this work, we conduct the first empirical comparative study of transfer learning for UATR, evaluating multiple pretrained audio models originating from diverse audio domains. The pretrained model weights are frozen, and the resulting embeddings are analyzed through classification, clustering, and similarity-based evaluations. The analysis shows that the geometrical structure of the embedding space is largely dominated by recording-specific characteristics. However, a simple linear probe can effectively suppress this recording-specific information and isolate ship-type features from these embeddings. As a result, linear probing enables effective automatic UATR using pretrained audio models at low computational cost, significantly reducing the need for a large amounts of high-quality labeled ship recordings.

Ranking models are extensively used in e-commerce for relevance estimation. These models often suffer from poor interpretability and no scale calibration, particularly when trained with typical ranking loss functions. This paper addresses the problem of post-hoc calibration of ranking models. We introduce MLPlatt: a simple yet effective ranking model calibration method that preserves the item ordering and converts ranker outputs to interpretable click-through rate (CTR) probabilities usable in downstream tasks. The method is context-aware by design and achieves good calibration metrics globally, and within strata corresponding to different values of a selected categorical field (such as user country or device), which is often important from a business perspective of an E-commerce platform. We demonstrate the superiority of MLPlatt over existing approaches on two datasets, achieving an improvement of over 10\% in F-ECE (Field Expected Calibration Error) compared to other methods. Most importantly, we show that high-quality calibration can be achieved without compromising the ranking quality.

Multi-agent LLM systems routinely generate multiple candidate responses that are aggregated by an LLM judge. To reduce the dominant prefill cost in such pipelines, recent work advocates KV cache reuse across partially shared contexts and reports substantial speedups for generation agents. In this work, we show that these efficiency gains do not transfer uniformly to judge-centric inference. Across GSM8K, MMLU, and HumanEval, we find that reuse strategies that are effective for execution agents can severely perturb judge behavior: end-task accuracy may appear stable, yet the judge's selection becomes highly inconsistent with dense prefill. We quantify this risk using Judge Consistency Rate (JCR) and provide diagnostics showing that reuse systematically weakens cross-candidate attention, especially for later candidate blocks. Our ablation further demonstrates that explicit cross-candidate interaction is crucial for preserving dense-prefill decisions. Overall, our results identify a previously overlooked failure mode of KV cache reuse and highlight judge-centric inference as a distinct regime that demands dedicated, risk-aware system design.

Mental manipulation, the strategic use of language to covertly influence or exploit others, is a newly emerging task in computational social reasoning. Prior work has focused exclusively on textual conversations, overlooking how manipulative tactics manifest in speech. We present the first study of mental manipulation detection in spoken dialogues, introducing a synthetic multi-speaker benchmark SPEECHMENTALMANIP that augments a text-based dataset with high-quality, voice-consistent Text-to-Speech rendered audio. Using few-shot large audio-language models and human annotation, we evaluate how modality affects detection accuracy and perception. Our results reveal that models exhibit high specificity but markedly lower recall on speech compared to text, suggesting sensitivity to missing acoustic or prosodic cues in training. Human raters show similar uncertainty in the audio setting, underscoring the inherent ambiguity of manipulative speech. Together, these findings highlight the need for modality-aware evaluation and safety alignment in multimodal dialogue systems.

Automated machine learning (AutoML) frameworks can lower technical barriers for predictive and prognostic model development in radiomics by enabling researchers without programming expertise to build models. However, their effectiveness in addressing radiomics-specific challenges remains unclear. This study evaluates the performance, efficiency, and accessibility of general-purpose and radiomics-specific AutoML frameworks on diverse radiomics classification tasks, thereby highlighting development needs for radiomics. Ten public/private radiomics datasets with varied imaging modalities (CT/MRI), sizes, anatomies and endpoints were used. Six general-purpose and five radiomics-specific frameworks were tested with predefined parameters using standardized cross-validation. Evaluation metrics included AUC, runtime, together with qualitative aspects related to software status, accessibility, and interpretability. Simplatab, a radiomics-specific tool with a no-code interface, achieved the highest average test AUC (81.81%) with a moderate runtime (~1 hour). LightAutoML, a general-purpose framework, showed the fastest execution with competitive performance (78.74% mean AUC in six minutes). Most radiomics-specific frameworks were excluded from the performance analysis due to obsolescence, extensive programming requirements, or computational inefficiency. Conversely, general-purpose frameworks demonstrated higher accessibility and ease of implementation. Simplatab provides an effective balance of performance, efficiency, and accessibility for radiomics classification problems. However, significant gaps remain, including the lack of accessible survival analysis support and the limited integration of feature reproducibility and harmonization within current AutoML frameworks. Future research should focus on adapting AutoML solutions to better address these radiomics-specific challenges.

LLM-based agent architectures systematically conflate information transport mechanisms with epistemic justification mechanisms. We formalize this class of architectural failures as semantic laundering: a pattern where propositions with absent or weak warrant are accepted by the system as admissible by crossing architecturally trusted interfaces. We show that semantic laundering constitutes an architectural realization of the Gettier problem: propositions acquire high epistemic status without a connection between their justification and what makes them true. Unlike classical Gettier cases, this effect is not accidental; it is architecturally determined and systematically reproducible. The central result is the Theorem of Inevitable Self-Licensing: under standard architectural assumptions, circular epistemic justification cannot be eliminated. We introduce the Warrant Erosion Principle as the fundamental explanation for this effect and show that scaling, model improvement, and LLM-as-judge schemes are structurally incapable of eliminating a problem that exists at the type level.

Generative Adversarial Networks (GANs) face a significant challenge of striking an optimal balance between high-quality image generation and training stability. Recent techniques, such as DCGAN, BigGAN, and StyleGAN, improve visual fidelity; however, such techniques usually struggle with mode collapse and unstable gradients at high network depth. This paper proposes a novel GAN structural model that incorporates deeper inception-inspired convolution and dilated convolution. This novel model is termed the Inception Generative Adversarial Network (IGAN). The IGAN model generates high-quality synthetic images while maintaining training stability, by reducing mode collapse as well as preventing vanishing and exploding gradients. Our proposed IGAN model achieves the Frechet Inception Distance (FID) of 13.12 and 15.08 on the CUB-200 and ImageNet datasets, respectively, representing a 28-33% improvement in FID over the state-of-the-art GANs. Additionally, the IGAN model attains an Inception Score (IS) of 9.27 and 68.25, reflecting improved image diversity and generation quality. Finally, the two techniques of dropout and spectral normalization are utilized in both the generator and discriminator structures to further mitigate gradient explosion and overfitting. These findings confirm that the IGAN model potentially balances training stability with image generation quality, constituting a scalable and computationally efficient framework for high-fidelity image synthesis.

Understanding and controlling the behavior of large language models (LLMs) is an increasingly important topic in multilingual NLP. Beyond prompting or fine-tuning, , i.e.,~manipulating internal representations during inference, has emerged as a more efficient and interpretable technique for adapting models to a target language. Yet, no dedicated benchmarks or evaluation protocols exist to quantify the effectiveness of steering techniques. We introduce CLaS-Bench, a lightweight parallel-question benchmark for evaluating language-forcing behavior in LLMs across 32 languages, enabling systematic evaluation of multilingual steering methods. We evaluate a broad array of steering techniques, including residual-stream DiffMean interventions, probe-derived directions, language-specific neurons, PCA/LDA vectors, Sparse Autoencoders, and prompting baselines. Steering performance is measured along two axes: language control and semantic relevance, combined into a single harmonic-mean steering score. We find that across languages simple residual-based DiffMean method consistently outperforms all other methods. Moreover, a layer-wise analysis reveals that language-specific structure emerges predominantly in later layers and steering directions cluster based on language family. CLaS-Bench is the first standardized benchmark for multilingual steering, enabling both rigorous scientific analysis of language representations and practical evaluation of steering as a low-cost adaptation alternative.

This paper introduces a decentralized multi-agent reinforcement learning framework enabling structurally heterogeneous teams of agents to jointly discover and acquire randomly located targets in environments characterized by partial observability, communication constraints, and dynamic interactions. Each agent's policy is trained with the Multi-Agent Proximal Policy Optimization algorithm and employs a Graph Attention Network encoder that integrates simulated range-sensing data with communication embeddings exchanged among neighboring agents, enabling context-aware decision-making from both local sensing and relational information. In particular, this work introduces a unified framework that integrates graph-based communication and trajectory-aware safety through safety filters. The architecture is supported by a structured reward formulation designed to encourage effective target discovery and acquisition, collision avoidance, and de-correlation between the agents' communication vectors by promoting informational orthogonality. The effectiveness of the proposed reward function is demonstrated through a comprehensive ablation study. Moreover, simulation results demonstrate safe and stable task execution, confirming the framework's effectiveness.

Equipping agents with memory is essential for solving real-world long-horizon problems. However, most existing agent memory mechanisms rely on static and hand-crafted workflows. This limits the performance and generalization ability of these memory designs, which highlights the need for a more flexible, learning-based memory framework. In this paper, we propose AtomMem, which reframes memory management as a dynamic decision-making problem. We deconstruct high-level memory processes into fundamental atomic CRUD (Create, Read, Update, Delete) operations, transforming the memory workflow into a learnable decision process. By combining supervised fine-tuning with reinforcement learning, AtomMem learns an autonomous, task-aligned policy to orchestrate memory behaviors tailored to specific task demands. Experimental results across 3 long-context benchmarks demonstrate that the trained AtomMem-8B consistently outperforms prior static-workflow memory methods. Further analysis of training dynamics shows that our learning-based formulation enables the agent to discover structured, task-aligned memory management strategies, highlighting a key advantage over predefined routines.

Background:Subject-specific finite element analysis accurately characterizes knee joint mechanics but is computationally expensive. Deep surrogate models provide a rapid alternative, yet their generalization across subjects under limited pose and load inputs remains unclear. It remains unclear whether the dominant source of prediction uncertainty arises from temporal history dependence or spatial propagation dependence. Methods:To disentangle these factors, we employed a shared MGN backbone with a fixed mesh topology. A dataset of running trials from nine subjects was constructed using an OpenSim-FEBio workflow. We developed four model variants to isolate specific dependencies: (1) a baseline MGN; (2) CT-MGN, incorporating a Control Transformer to encode short-horizon history; (3) MsgModMGN, applying state-conditioned modulation to message passing for adaptive propagation; (4) CT-MsgModMGN, combining both mechanisms. Models were evaluated using a rigorous grouped 3-fold cross-validation on unseen subjects.Results:The models incorporating history encoding significantly outperformed the baseline MGN and MsgModMGN in global accuracy and spatial consistency. Crucially, the CT module effectively mitigated the peak-shaving defect common in deep surrogates, significantly reducing peak stress prediction errors. In contrast, the spatial propagation modulation alone yielded no significant improvement over the baseline, and combining it with CT provided no additional benefit.Conclusion:Temporal history dependence, rather than spatial propagation modulation, is the primary driver of prediction uncertainty in cross-subject knee contact mechanics. Explicitly encoding short-horizon driver sequences enables the surrogate model to recover implicit phase information, thereby achieving superior fidelity in peak-stress capture and high-risk localization compared to purely state-based approaches.

Deep double descent is one of the key phenomena underlying the generalization capability of deep learning models. In this study, epoch-wise double descent, which is delayed generalization following overfitting, was empirically investigated by focusing on the evolution of internal structures. Fully connected neural networks of three different sizes were trained on the CIFAR-10 dataset with 30% label noise. By decomposing the loss curves into signal contributions from clean and noisy training data, the epoch-wise evolutions of internal signals were analyzed separately. Three main findings were obtained from this analysis. First, the model achieved strong re-generalization on test data even after perfectly fitting noisy training data during the double descent phase, corresponding to a "benign overfitting" state. Second, noisy data were learned after clean data, and as learning progressed, their corresponding internal activations became increasingly separated in outer layers; this enabled the model to overfit only noisy data. Third, a single, very large activation emerged in the shallow layer across all models; this phenomenon is referred as "outliers," "massive activa-tions," and "super activations" in recent large language models and evolves with re-generalization. The magnitude of large activation correlated with input patterns but not with output patterns. These empirical findings directly link the recent key phenomena of "deep double descent," "benign overfitting," and "large activation", and support the proposal of a novel scenario for understanding deep double descent.

Large Multimodal Models (LMMs) have recently shown remarkable promise in low-level visual perception tasks, particularly in Image Quality Assessment (IQA), demonstrating strong zero-shot capability. However, achieving state-of-the-art performance often requires computationally expensive fine-tuning methods, which aim to align the distribution of quality-related token in output with image quality levels. Inspired by recent training-free works for LMM, we introduce IQARAG, a novel, training-free framework that enhances LMMs' IQA ability. IQARAG leverages Retrieval-Augmented Generation (RAG) to retrieve some semantically similar but quality-variant reference images with corresponding Mean Opinion Scores (MOSs) for input image. These retrieved images and input image are integrated into a specific prompt. Retrieved images provide the LMM with a visual perception anchor for IQA task. IQARAG contains three key phases: Retrieval Feature Extraction, Image Retrieval, and Integration & Quality Score Generation. Extensive experiments across multiple diverse IQA datasets, including KADID, KonIQ, LIVE Challenge, and SPAQ, demonstrate that the proposed IQARAG effectively boosts the IQA performance of LMMs, offering a resource-efficient alternative to fine-tuning for quality assessment.

Recent Large Reasoning Models (LRMs) achieve strong performance by leveraging long-form Chain-of-Thought (CoT) reasoning, but uniformly applying overlong reasoning at inference time incurs substantial and often unnecessary computational cost. To address this, prior work explores various strategies to infer an appropriate reasoning budget from the input. However, such approaches are unreliable in the worst case, as estimating the minimal required reasoning effort is fundamentally difficult, and they implicitly fix the trade-off between reasoning cost and accuracy during training, limiting flexibility under varying deployment scenarios. Motivated by these limitations, we propose ORBIT, a controllable multi-budget reasoning framework with well-separated reasoning modes triggered by input. ORBIT employs multi-stage reinforcement learning to discover Pareto-optimal reasoning behaviors at each effort, followed by on-policy distillation to fuse these behaviors into a single unified model. Experiments show that ORBIT achieves (1) controllable reasoning behavior over multiple modes, (2) competitive reasoning density within each mode, and (3) integration of these frontier policies into a single unified student model while preserving clear mode separation and high per-mode performance.

Intelligent agent systems in real-world agricultural scenarios must handle diverse tasks under multimodal inputs, ranging from lightweight information understanding to complex multi-step execution. However, most existing approaches rely on a unified execution paradigm, which struggles to accommodate large variations in task complexity and incomplete tool availability commonly observed in agricultural environments. To address this challenge, we propose AgriAgent, a two-level agent framework for real-world agriculture. AgriAgent adopts a hierarchical execution strategy based on task complexity: simple tasks are handled through direct reasoning by modality-specific agents, while complex tasks trigger a contract-driven planning mechanism that formulates tasks as capability requirements and performs capability-aware tool orchestration and dynamic tool generation, enabling multi-step and verifiable execution with failure recovery. Experimental results show that AgriAgent achieves higher execution success rates and robustness on complex tasks compared to existing tool-centric agent baselines that rely on unified execution paradigms. All code, data will be released at after our work be accepted to promote reproducible research.

This study investigates the use of prompt engineering to enhance large language models (LLMs), specifically GPT-4o-mini and gemini-1.5-flash, in sentiment analysis tasks. It evaluates advanced prompting techniques like few-shot learning, chain-of-thought prompting, and self-consistency against a baseline. Key tasks include sentiment classification, aspect-based sentiment analysis, and detecting subtle nuances such as irony. The research details the theoretical background, datasets, and methods used, assessing performance of LLMs as measured by accuracy, recall, precision, and F1 score. Findings reveal that advanced prompting significantly improves sentiment analysis, with the few-shot approach excelling in GPT-4o-mini and chain-of-thought prompting boosting irony detection in gemini-1.5-flash by up to 46%. Thus, while advanced prompting techniques overall improve performance, the fact that few-shot prompting works best for GPT-4o-mini and chain-of-thought excels in gemini-1.5-flash for irony detection suggests that prompting strategies must be tailored to both the model and the task. This highlights the importance of aligning prompt design with both the LLM's architecture and the semantic complexity of the task.

Large Language Models (LLMs) often exhibit slash attention patterns, where attention scores concentrate along the $Δ$-th sub-diagonal for some offset $Δ$. These patterns play a key role in passing information across tokens. But why do they emerge? In this paper, we demystify the emergence of these Slash-Dominant Heads (SDHs) from both empirical and theoretical perspectives. First, by analyzing open-source LLMs, we find that SDHs are intrinsic to models and generalize to out-of-distribution prompts. To explain the intrinsic emergence, we analyze the queries, keys, and Rotary Position Embedding (RoPE), which jointly determine attention scores. Our empirical analysis reveals two characteristic conditions of SDHs: (1) Queries and keys are almost rank-one, and (2) RoPE is dominated by medium- and high-frequency components. Under these conditions, queries and keys are nearly identical across tokens, and interactions between medium- and high-frequency components of RoPE give rise to SDHs. Beyond empirical evidence, we theoretically show that these conditions are sufficient to ensure the emergence of SDHs by formalizing them as our modeling assumptions. Particularly, we analyze the training dynamics of a shallow Transformer equipped with RoPE under these conditions, and prove that models trained via gradient descent exhibit SDHs. The SDHs generalize to out-of-distribution prompts.

Chinese stand-up comedy generation goes beyond plain text generation, requiring culturally grounded humor, precise timing, stage-performance cues, and implicit multi-step reasoning. Moreover, commonly used Chinese humor datasets are often better suited for humor understanding and evaluation than for long-form stand-up generation, making direct supervision misaligned with the target task. To address these challenges, we present OpenMic, an end-to-end multi-agent system built on AutoGen that transforms a user-provided life topic into a 3-5 minute Chinese stand-up performance and further produces a narrated comedy video. OpenMic orchestrates multiple specialized agents in a multi-round iterative loop-planning to jointly optimize humor, timing, and performability. To mitigate the dataset-task mismatch, we augment generation with retrieval-augmented generation (RAG) for material grounding and idea expansion, and we fine-tune a dedicated JokeWriter to better internalize stand-up-specific setup-punchline structures and long-range callbacks.

As the volume of unstructured text continues to grow across domains, there is an urgent need for scalable methods that enable interpretable organization, summarization, and retrieval of information. This work presents a unified framework for interpretable topic modeling, zero-shot topic labeling, and topic-guided semantic retrieval over large agricultural text corpora. Leveraging BERTopic, we extract semantically coherent topics. Each topic is converted into a structured prompt, enabling a language model to generate meaningful topic labels and summaries in a zero-shot manner. Querying and document exploration are supported via dense embeddings and vector search, while a dedicated evaluation module assesses topical coherence and bias. This framework supports scalable and interpretable information access in specialized domains where labeled data is limited.

Recent search-augmented LLMs trained with reinforcement learning (RL) can interleave searching and reasoning for multi-hop reasoning tasks. However, they face two critical failure modes as the accumulating context becomes flooded with both crucial evidence and irrelevant information: (1) ineffective search chain construction that produces incorrect queries or omits retrieval of critical information, and (2) reasoning hijacking by peripheral evidence that causes models to misidentify distractors as valid evidence. To address these challenges, we propose **D$^2$Plan**, a **D**ual-agent **D**ynamic global **Plan**ning paradigm for complex retrieval-augmented reasoning. **D$^2$Plan** operates through the collaboration of a *Reasoner* and a *Purifier*: the *Reasoner* constructs explicit global plans during reasoning and dynamically adapts them based on retrieval feedback; the *Purifier* assesses retrieval relevance and condenses key information for the *Reasoner*. We further introduce a two-stage training framework consisting of supervised fine-tuning (SFT) cold-start on synthesized trajectories and RL with plan-oriented rewards to teach LLMs to master the **D$^2$Plan** paradigm. Extensive experiments demonstrate that **D$^2$Plan** enables more coherent multi-step reasoning and stronger resilience to irrelevant information, thereby achieving superior performance on challenging QA benchmarks.

Modern agentic systems operate in environments with extremely large action spaces, such as tool-augmented language models with thousands of available APIs or retrieval operations. Despite this scale, empirical evidence suggests that only a small subset of actions meaningfully influences performance in a given deployment. Motivated by this observation, we study a contextual linear reward model in which action relevance is governed by a structured sparsity assumption: only a small number of actions have nonzero effects across latent states. We formulate action discovery as a block-sparse recovery problem and analyze a greedy algorithm inspired by Orthogonal Matching Pursuit. Under standard assumptions on incoherence, signal strength, and action coverage, we prove that the greedy procedure exactly recovers the relevant action set with high probability, using a number of samples that scales polynomially in the sparsity level and latent dimension, and only logarithmically in the total number of actions. We further provide estimation error guarantees for refitted parameters and show that the resulting decision rule is near-optimal for new latent states. Complementing these results, we establish information-theoretic lower bounds demonstrating that sparsity and sufficient coverage are necessary for tractability. Together, our results identify sparse action discovery as a fundamental principle underlying large-action decision-making and provide a theoretical foundation for action pruning in agentic systems.

Convolutional neural networks (CNNs) have been widely used in the computer vision community, significantly improving the state-of-the-art. But learning good features often is computationally expensive in machine learning settings and is especially difficult when there is a lack of data. One-shot learning is one such area where only limited data is available. In one-shot learning, predictions have to be made after seeing only one example from one class, which requires special techniques. In this paper we explore different approaches to one-shot identification tasks in different domains including an industrial application and face recognition. We use a special technique with stacked images and use siamese capsule networks. It is encouraging to see that the approach using capsule architecture achieves strong results and exceeds other techniques on a wide range of datasets from industrial application to face recognition benchmarks while being easy to use and optimise.

Particle-in-Cell (PIC) simulations spend most of their execution time on particle--grid interactions, where fine-grained atomic updates become a major bottleneck on traditional many-core CPUs. Recent CPU architectures integrate specialized Matrix Processing Units (MPUs) that efficiently support matrix outer-product operations, offering new opportunities to overcome this limitation. Leveraging this architectural shift, this work focuses on redesigning the current deposition step of PIC simulations under a matrix-centric execution model. We present MatrixPIC, the first holistic co-design of the deposition kernel, data layout, and incremental particle sorting tailored to the hybrid MPU--VPU SIMD model on modern CPUs. MatrixPIC introduces: (i)~a block-matrix formulation of the current deposition algorithm that maps naturally to MPU outer-product primitives; (ii)~a hybrid execution pipeline that combines MPU-based high-density accumulation with VPU-based data preparation and control flow; and (iii)~an $O(1)$-amortized incremental sorter based on a gapped packed-memory array to preserve data locality for efficient MPU execution. Evaluated on a next-generation HPC platform, MatrixPIC achieves significant performance gains. In Laser-Wakefield Acceleration (LWFA) simulations, it delivers up to $2.63\times$ speedup in total runtime. For third-order deposition, the core kernel is accelerated by $8.7\times$ over the baseline and $2.0\times$ over the best hand-optimized VPU implementation. Moreover, MatrixPIC reaches $83.08\%$ of theoretical CPU peak performance, nearly $2.8\times$ higher than a highly optimized CUDA kernel on a data center GPU. These results demonstrate the effectiveness of matrix-oriented co-design for accelerating PIC simulations on emerging CPU architectures.

With the rise of the Agent Web and Model Context Protocol (MCP), the agent ecosystem is evolving into an open collaborative network, exponentially increasing accessible tools. However, current architectures face severe scalability and generality bottlenecks. To address this, we propose ToolACE-MCP, a pipeline for training history-aware routers to empower precise navigation in large-scale ecosystems. By leveraging a dependency-rich candidate Graph to synthesize multi-turn trajectories, we effectively train routers with dynamic context understanding to create the plug-and-play Light Routing Agent. Experiments on the real-world benchmarks MCP-Universe and MCP-Mark demonstrate superior performance. Notably, ToolACE-MCP exhibits critical properties for the future Agent Web: it not only generalizes to multi-agent collaboration with minimal adaptation but also maintains exceptional robustness against noise and scales effectively to massive candidate spaces. These findings provide a strong empirical foundation for universal orchestration in open-ended ecosystems.

Tool-Integrated Reasoning has emerged as a key paradigm to augment Large Language Models (LLMs) with computational capabilities, yet integrating tool-use into long Chain-of-Thought (long CoT) remains underexplored, largely due to the scarcity of training data and the challenge of integrating tool-use without compromising the model's intrinsic long-chain reasoning. In this paper, we introduce DART (Discovery And Reinforcement of Tool-Integrated Reasoning Chains via Rollout Trees), a reinforcement learning framework that enables spontaneous tool-use during long CoT reasoning without human annotation. DART operates by constructing dynamic rollout trees during training to discover valid tool-use opportunities, branching out at promising positions to explore diverse tool-integrated trajectories. Subsequently, a tree-based process advantage estimation identifies and credits specific sub-trajectories where tool invocation positively contributes to the solution, effectively reinforcing these beneficial behaviors. Extensive experiments on challenging benchmarks like AIME and GPQA-Diamond demonstrate that DART significantly outperforms existing methods, successfully harmonizing tool execution with long CoT reasoning.

Speculative decoding (SD) has emerged as a promising approach to accelerate LLM inference without sacrificing output quality. Existing SD methods tailored for video-LLMs primarily focus on pruning redundant visual tokens to mitigate the computational burden of massive visual inputs. However, existing methods do not achieve inference acceleration comparable to text-only LLMs. We observe from extensive experiments that this phenomenon mainly stems from two limitations: (i) their pruning strategies inadequately preserve visual semantic tokens, degrading draft quality and acceptance rates; (ii) even with aggressive pruning (e.g., 90% visual tokens removed), the draft model's remaining inference cost limits overall speedup. To address these limitations, we propose HIPPO, a general holistic-aware parallel speculative decoding framework. Specifically, HIPPO proposes (i) a semantic-aware token preservation method, which fuses global attention scores with local visual semantics to retain semantic information at high pruning ratios; (ii) a video parallel SD algorithm that decouples and overlaps draft generation and target verification phases. Experiments on four video-LLMs across six benchmarks demonstrate HIPPO's effectiveness, yielding up to 3.51x speedup compared to vanilla auto-regressive decoding.

Tool-augmented LLM systems expose a control regime that learning theory has largely ignored: sequential decision-making with a massive discrete action universe (tools, APIs, documents) in which only a small, unknown subset is relevant for any fixed task distribution. We formalize this setting as Sparse Agentic Control (SAC), where policies admit block-sparse representations over M >> 1 actions and rewards depend on sparse main effects and (optionally) sparse synergies. We study ell_{1,2}-regularized policy learning through a convex surrogate and establish sharp, compressed-sensing-style results: (i) estimation and value suboptimality scale as k (log M / T)^{1/2} under a Policy-RSC condition; (ii) exact tool-support recovery holds via primal-dual witness arguments when T > k log M under incoherence and beta-min; and (iii) any dense policy class requires Omega(M) samples, explaining the instability of prompt-only controllers. We further show that under partial observability, LLMs matter only through a belief/representation error epsilon_b, yielding an additive O(epsilon_b) degradation while preserving logarithmic dependence on M. Extensions cover tuning-free, online, robust, group-sparse, and interaction-aware SAC.

While reasoning-enhanced large language models perform strongly on English medical tasks, a persistent multilingual gap remains, with substantially weaker reasoning in local languages, limiting equitable global medical deployment. To bridge this gap, we introduce Med-CoReasoner, a language-informed co-reasoning framework that elicits parallel English and local-language reasoning, abstracts them into structured concepts, and integrates local clinical knowledge into an English logical scaffold via concept-level alignment and retrieval. This design combines the structural robustness of English reasoning with the practice-grounded expertise encoded in local languages. To evaluate multilingual medical reasoning beyond multiple-choice settings, we construct MultiMed-X, a benchmark covering seven languages with expert-annotated long-form question answering and natural language inference tasks, comprising 350 instances per language. Experiments across three benchmarks show that Med-CoReasoner improves multilingual reasoning performance by an average of 5%, with particularly substantial gains in low-resource languages. Moreover, model distillation and expert evaluation analysis further confirm that Med-CoReasoner produces clinically sound and culturally grounded reasoning traces.

Hand gesture recognition is an important aspect of human-computer interaction. It forms the basis of sign language for the visually impaired people. This work proposes a novel hand gesture recognizing system for the differently-abled persons. The model uses a convolutional neural network, known as VGG-16 net, for building a trained model on a widely used image dataset by employing Python and Keras libraries. Furthermore, the result is validated by the NUS dataset, consisting of 10 classes of hand gestures, fed to the model as the validation set. Afterwards, a testing dataset of 10 classes is built by employing Google's open source Application Programming Interface (API) that captures different gestures of human hand and the efficacy is then measured by carrying out experiments. The experimental results show that by combining a transfer learning mechanism together with the image data augmentation, the VGG-16 net produced around 98% accuracy.

Cardiac amyloidosis (CA) is a rare and underdiagnosed infiltrative cardiomyopathy, and available datasets for machine-learning models are typically small, imbalanced and heterogeneous. This paper presents a Generative Adversarial Network (GAN) and a graphical command-line interface for generating realistic synthetic electrocardiogram (ECG) beats to support early diagnosis and patient stratification in CA. The tool is designed for usability, allowing clinical researchers to train class-specific generators once and then interactively produce large volumes of labelled synthetic beats that preserve the distribution of minority classes.

We introduce T3 (Testing Trustworthy Thinking), a diagnostic benchmark designed to rigorously evaluate LLM causal judgment across Pearl's Ladder of Causality. Comprising 454 expert-curated vignettes, T3 prioritizes high-resolution failure analysis, decomposing performance into Utility (sensitivity), Safety (specificity), and Wise Refusal on underdetermined cases. By applying T3 to frontier models, we diagnose two distinct pathologies: a "Skepticism Trap" at L1 (where safety-tuned models like Claude Haiku reject 60% of valid links) and a non-monotonic Scaling Paradox at L3. In the latter, the larger GPT-5.2 underperforms GPT-4-Turbo by 55 points on ambiguous counterfactuals, driven by a collapse into paralysis (excessive hedging) rather than hallucination. Finally, we use the benchmark to validate a process-verified protocol (RCA), showing that T3 successfully captures the restoration of decisive causal judgment under structured verification.

Unsupervised feature selection aims to identify a compact subset of features that captures the intrinsic structure of data without supervised label. Most existing studies evaluate the performance of methods using the single-label dataset that can be instantiated by selecting a label from multi-label data while maintaining the original features. Because the chosen label can vary arbitrarily depending on the experimental setting, the superiority among compared methods can be changed with regard to which label happens to be selected. Thus, evaluating unsupervised feature selection methods based solely on single-label accuracy is unreasonable for assessing their true discriminative ability. This study revisits this evaluation paradigm by adopting a multi-label classification framework. Experiments on 21 multi-label datasets using several representative methods demonstrate that performance rankings differ markedly from those reported under single-label settings, suggesting the possibility of multi-label evaluation settings for fair and reliable comparison of unsupervised feature selection methods.

Large AI Model (LAM) have been proposed to applications of Non-Terrestrial Networks (NTN), that offer better performance with its great generalization and reduced task specific trainings. In this paper, we propose a Deep Reinforcement Learning (DRL) agent that is guided by a Large Language Model (LLM). The LLM operates as a high level coordinator that generates textual guidance that shape the reward of the DRL agent during training. The results show that the LAM-DRL outperforms the traditional DRL by 40% in nominal weather scenarios and 64% in extreme weather scenarios compared to heuristics in terms of throughput, fairness, and outage probability.

We analyze initialization dynamics for LDLT-based $\mathcal{L}$-Lipschitz layers by deriving the exact marginal output variance when the underlying parameter matrix $W_0\in \mathbb{R}^{m\times n}$ is initialized with IID Gaussian entries $\mathcal{N}(0,σ^2)$. The Wishart distribution, $S=W_0W_0^\top\sim\mathcal{W}_m(n,σ^2 \boldsymbol{I}_m)$, used for computing the output marginal variance is derived in closed form using expectations of zonal polynomials via James' theorem and a Laplace-integral expansion of $(α\boldsymbol{I}_m+S)^{-1}$. We develop an Isserlis/Wick-based combinatorial expansion for $\operatorname{\mathbb{E}}\left[\operatorname{tr}(S^k)\right]$ and provide explicit truncated moments up to $k=10$, which yield accurate series approximations for small-to-moderate $σ^2$. Monte Carlo experiments confirm the theoretical estimates. Furthermore, empirical analysis was performed to quantify that, using current He or Kaiming initialization with scaling $1/\sqrt{n}$, the output variance is $0.41$, whereas the new parameterization with $10/ \sqrt{n}$ for $α=1$ results in an output variance of $0.9$. The findings clarify why deep $\mathcal{L}$-Lipschitz networks suffer rapid information loss at initialization and offer practical prescriptions for choosing initialization hyperparameters to mitigate this effect. However, using the Higgs boson classification dataset, a hyperparameter sweep over optimizers, initialization scale, and depth was conducted to validate the results on real-world data, showing that although the derivation ensures variance preservation, empirical results indicate He initialization still performs better.

In heterogeneous graphs, we can observe complex structures such as tree-like or hierarchical structures. Recently, the hyperbolic space has been widely adopted in many studies to effectively learn these complex structures. Although these methods have demonstrated the advantages of the hyperbolic space in learning heterogeneous graphs, most existing methods still have several challenges. They rely heavily on tangent-space operations, which often lead to mapping distortions during frequent transitions. Moreover, their message-passing architectures mainly focus on local neighborhood information, making it difficult to capture global hierarchical structures and long-range dependencies between different types of nodes. To address these limitations, we propose Hyperbolic Heterogeneous Graph Transformer (HypHGT), which effectively and efficiently learns heterogeneous graph representations entirely within the hyperbolic space. Unlike previous message-passing based hyperbolic heterogeneous GNNs, HypHGT naturally captures both local and global dependencies through transformer-based architecture. Furthermore, the proposed relation-specific hyperbolic attention mechanism in HypHGT, which operates with linear time complexity, enables efficient computation while preserving the heterogeneous information across different relation types. This design allows HypHGT to effectively capture the complex structural properties and semantic information inherent in heterogeneous graphs. We conduct comprehensive experiments to evaluate the effectiveness and efficiency of HypHGT, and the results demonstrate that it consistently outperforms state-of-the-art methods in node classification task, with significantly reduced training time and memory usage.

Financial markets are influenced by human behavior that deviates from rationality due to cognitive biases. Traditional reinforcement learning (RL) models for financial decision-making assume rational agents, potentially overlooking the impact of psychological factors. This study integrates cognitive biases into RL frameworks for financial trading, hypothesizing that such models can exhibit human-like trading behavior and achieve better risk-adjusted returns than standard RL agents. We introduce biases, such as overconfidence and loss aversion, into reward structures and decision-making processes and evaluate their performance in simulated and real-world trading environments. Despite its inconclusive or negative results, this study provides insights into the challenges of incorporating human-like biases into RL, offering valuable lessons for developing robust financial AI systems.

Unobtrusive sensor-based recognition of Activities of Daily Living (ADLs) in smart homes by processing data collected from IoT sensing devices supports applications such as healthcare, safety, and energy management. Recent zero-shot methods based on Large Language Models (LLMs) have the advantage of removing the reliance on labeled ADL sensor data. However, existing approaches rely on time-based segmentation, which is poorly aligned with the contextual reasoning capabilities of LLMs. Moreover, existing approaches lack methods for estimating prediction confidence. This paper proposes to improve zero-shot ADL recognition with event-based segmentation and a novel method for estimating prediction confidence. Our experimental evaluation shows that event-based segmentation consistently outperforms time-based LLM approaches on complex, realistic datasets and surpasses supervised data-driven methods, even with relatively small LLMs (e.g., Gemma 3 27B). The proposed confidence measure effectively distinguishes correct from incorrect predictions.

Reward engineering, the manual specification of reward functions to induce desired agent behavior, remains a fundamental challenge in multi-agent reinforcement learning. This difficulty is amplified by credit assignment ambiguity, environmental non-stationarity, and the combinatorial growth of interaction complexity. We argue that recent advances in large language models (LLMs) point toward a shift from hand-crafted numerical rewards to language-based objective specifications. Prior work has shown that LLMs can synthesize reward functions directly from natural language descriptions (e.g., EUREKA) and adapt reward formulations online with minimal human intervention (e.g., CARD). In parallel, the emerging paradigm of Reinforcement Learning from Verifiable Rewards (RLVR) provides empirical evidence that language-mediated supervision can serve as a viable alternative to traditional reward engineering. We conceptualize this transition along three dimensions: semantic reward specification, dynamic reward adaptation, and improved alignment with human intent, while noting open challenges related to computational overhead, robustness to hallucination, and scalability to large multi-agent systems. We conclude by outlining a research direction in which coordination arises from shared semantic representations rather than explicitly engineered numerical signals.

This work introduces a novel technique, named structural dimension reduction, to collapse a Bayesian network onto a minimum and localized one while ensuring that probabilistic inferences between the original and reduced networks remain consistent. To this end, we propose a new combinatorial structure in directed acyclic graphs called the directed convex hull, which has turned out to be equivalent to their minimum localized Bayesian networks. An efficient polynomial-time algorithm is devised to identify them by determining the unique directed convex hulls containing the variables of interest from the original networks. Experiments demonstrate that the proposed technique has high dimension reduction capability in real networks, and the efficiency of probabilistic inference based on directed convex hulls can be significantly improved compared with traditional methods such as variable elimination and belief propagation algorithms. The code of this study is open at \href{https://github.com/Balance-H/Algorithms}{https://github.com/Balance-H/Algorithms} and the proofs of the results in the main body are postponed to the appendix.

As language-model agents evolve from passive chatbots into proactive assistants that handle personal data, evaluating their adherence to social norms becomes increasingly critical, often through the lens of Contextual Integrity (CI). However, existing CI benchmarks are largely text-centric and primarily emphasize negative refusal scenarios, overlooking multimodal privacy risks and the fundamental trade-off between privacy and utility. In this paper, we introduce MPCI-Bench, the first Multimodal Pairwise Contextual Integrity benchmark for evaluating privacy behavior in agentic settings. MPCI-Bench consists of paired positive and negative instances derived from the same visual source and instantiated across three tiers: normative Seed judgments, context-rich Story reasoning, and executable agent action Traces. Data quality is ensured through a Tri-Principle Iterative Refinement pipeline. Evaluations of state-of-the-art multimodal models reveal systematic failures to balance privacy and utility and a pronounced modality leakage gap, where sensitive visual information is leaked more frequently than textual information. We will open-source MPCI-Bench to facilitate future research on agentic CI.

While Graph Neural Networks (GNNs) excel on graph-structured data, their performance is fundamentally limited by the quality of the observed graph, which often contains noise, missing links, or structural properties misaligned with GNNs' underlying assumptions. To address this, graph structure learning aims to infer a more optimal topology. Existing methods, however, often incur high computational costs due to complex generative models and iterative joint optimization, limiting their practical utility. In this paper, we propose GADPN, a simple yet effective graph structure learning framework that adaptively refines graph topology via low-rank denoising and generalized structural perturbation. Our approach makes two key contributions: (1) we introduce Bayesian optimization to adaptively determine the optimal denoising strength, tailoring the process to each graph's homophily level; and (2) we extend the structural perturbation method to arbitrary graphs via Singular Value Decomposition (SVD), overcoming its original limitation to symmetric structures. Extensive experiments on benchmark datasets demonstrate that GADPN achieves state-of-the-art performance while significantly improving efficiency. It shows particularly strong gains on challenging disassortative graphs, validating its ability to robustly learn enhanced graph structures across diverse network types.

This research analyzed and compared the multi-modal approach in the Vision Transformer(EVA-ViT) based image encoder with the LlaMA or ChatGPT LLM to reduce the hallucination problem and detect diseases in chest x-ray images. In this research, we utilized the NIH Chest X-ray image to train the model and compared it in image-based RAG, text-based RAG, and baseline. [3] [5] In a result, the text-based RAG[2] e!ectively reduces the hallucination problem by using external knowledge information, and the image-based RAG improved the prediction con"dence and calibration by using the KNN methods. [4] Moreover, the GPT LLM showed better performance, a low hallucination rate, and better Expected Calibration Error(ECE) than Llama Llama-based model. This research shows the challenge of data imbalance, a complex multi-stage structure, but suggests a large experience environment and a balanced example of use.

The recent paradigm shift toward large reasoning models (LRMs) as autonomous agents has intensified the demand for sophisticated, multi-turn tool-use capabilities. Yet, existing datasets and data-generation approaches are limited by static, predefined toolsets that cannot scale to the complexity of open-ended human-agent collaboration. To address this, we initially developed a framework for automated task-oriented multi-turn dialogue generation at scale, utilizing an LRM-based simulator to dynamically generate high-value, domain-specific tools to solve specified tasks. However, we observe that a purely task-oriented design often results in "solely task-solving" trajectories, where the agent completes the objective with minimal interaction, failing to generate the high turn-count conversations seen in realistic scenarios. To bridge this gap, we shift toward a user-oriented simulation paradigm. By decoupling task generation from a dedicated user simulator that mimics human behavioral rules - such as incremental request-making and turn-by-turn feedback - we facilitate more authentic, extended multi-turn dialogues that reflect the iterative nature of real-world problem solving. Our generation pipeline operates as a versatile, plug-and-play module capable of initiating generation from any state, ensuring high scalability in producing extended tool-use data. Furthermore, by facilitating multiple task completions within a single trajectory, it yields a high-density dataset that reflects the multifaceted demands of real-world human-agent interaction.

General intelligence must reorganize experience into internal structures that enable prediction and action under finite resources. Existing systems implicitly presuppose fixed primitive units -- tokens, subwords, pixels, or predefined sensor channels -- thereby bypassing the question of how representational units themselves emerge and stabilize. This paper proposes SANC(E3), an axiomatic framework in which representational units are not given a priori but instead arise as stable outcomes of competitive selection, reconstruction, and compression under finite activation capacity, governed by the explicit minimization of an energy functional E3. SANC(E3) draws a principled distinction between system tokens -- structural anchors such as {here, now, I} and sensory sources -- and tokens that emerge through self-organization during co-occurring events. Five core axioms formalize finite capacity, association from co-occurrence, similarity-based competition, confidence-based stabilization, and the reconstruction-compression-update trade-off. A key feature is a pseudo-memory-mapped I/O mechanism, through which internally replayed Gestalts are processed via the same axiomatic pathway as external sensory input. As a result, perception, imagination, prediction, planning, and action are unified within a single representational and energetic process. From the axioms, twelve propositions are derived, showing that category formation, hierarchical organization, unsupervised learning, and high-level cognitive activities can all be understood as instances of Gestalt completion under E3 minimization.

The rapid growth of large language models raises pressing concerns about intellectual property protection under black-box deployment. Existing backdoor-based fingerprints either rely on rare tokens -- leading to high-perplexity inputs susceptible to filtering -- or use fixed trigger-response mappings that are brittle to leakage and post-hoc adaptation. We propose \textsc{Dual-Layer Nested Fingerprinting} (DNF), a black-box method that embeds a hierarchical backdoor by coupling domain-specific stylistic cues with implicit semantic triggers. Across Mistral-7B, LLaMA-3-8B-Instruct, and Falcon3-7B-Instruct, DNF achieves perfect fingerprint activation while preserving downstream utility. Compared with existing methods, it uses lower-perplexity triggers, remains undetectable under fingerprint detection attacks, and is relatively robust to incremental fine-tuning and model merging. These results position DNF as a practical, stealthy, and resilient solution for LLM ownership verification and intellectual property protection.

Federated learning protocols require repeated synchronization between clients and a central server, with convergence rates depending on learning rates, data heterogeneity, and client sampling. This paper asks whether iterative communication is necessary for distributed linear regression. We show it is not. We formulate federated ridge regression as a distributed equilibrium problem where each client computes local sufficient statistics -- the Gram matrix and moment vector -- and transmits them once. The server reconstructs the global solution through a single matrix inversion. We prove exact recovery: under a coverage condition on client feature matrices, one-shot aggregation yields the centralized ridge solution, not an approximation. For heterogeneous distributions violating coverage, we derive non-asymptotic error bounds depending on spectral properties of the aggregated Gram matrix. Communication reduces from $\mathcal{O}(Rd)$ in iterative methods to $\mathcal{O}(d^2)$ total; for high-dimensional settings, we propose and experimentally validate random projection techniques reducing this to $\mathcal{O}(m^2)$ where $m \ll d$. We establish differential privacy guarantees where noise is injected once per client, eliminating the composition penalty that degrades privacy in multi-round protocols. We further address practical considerations including client dropout robustness, federated cross-validation for hyperparameter selection, and comparison with gradient-based alternatives. Comprehensive experiments on synthetic heterogeneous regression demonstrate that one-shot fusion matches FedAvg accuracy while requiring up to $38\times$ less communication. The framework applies to kernel methods and random feature models but not to general nonlinear architectures.

Modern Mixture-of-Experts (MoE) language models are designed based on total parameters (memory footprint) and active parameters (inference cost). However, we find these two factors alone are insufficient to describe an optimal architecture. Through a systematic study, we demonstrate that MoE performance is primarily determined by total parameters ($N_{total}$) and expert sparsity ($s:=n_{exp}/n_{topk}$). Moreover, $n_{exp}$ and $n_{topk}$ do not "cancel out" within the sparsity ratio; instead, a larger total number of experts slightly penalizes performance by forcing a reduction in core model dimensions (depth and width) to meet memory constraints. This motivates a simple principle for MoE design which maximizes $N_{total}$ while minimizing $s$ (maximizing $n_{topk}$) and $n_{exp}$ under the given constraints. Our findings provide a robust framework for resolving architectural ambiguity and guiding MoE design.

As one of the worldwide spread traditional game, Official International Mahjong can be played and promoted online through remote devices instead of requiring face-to-face interaction. However, online players have fragmented playtime and unfixed combination of opponents in contrary to offline players who have fixed opponents for multiple rounds of play. Therefore, the rules designed for offline players need to be modified to ensure the fairness of online single-round play. Specifically, We employ a world champion AI to engage in self-play competitions and conduct statistical data analysis. Our study reveals the first-mover advantage and issues in the subgoal scoring settings. Based on our findings, we propose rule adaptations to make the game more suitable for the online environment, such as introducing compensatory points for the first-mover advantage and refining the scores of subgoals for different tile patterns. Compared with the traditional method of rotating positions over multiple rounds to balance first-mover advantage, our compensatory points mechanism in each round is more convenient for online players. Furthermore, we implement the revised Mahjong game online, which is open for online players. This work is an initial attempt to use data from AI systems to evaluate Official Internatinoal Mahjong's game balance and develop a revised version of the traditional game better adapted for online players.

Multiagent reinforcement learning (MARL) has attracted considerable attention due to its potential in addressing complex cooperative tasks. However, existing MARL approaches often rely on frequent exchanges of action or state information among agents to achieve effective coordination, which is difficult to satisfy in practical robotic systems. A common solution is to introduce estimator networks to model the behaviors of other agents and predict their actions; nevertheless, such designs cause the size and computational cost of the estimator networks to grow rapidly with the number of agents, thereby limiting scalability in large-scale systems. To address these challenges, this paper proposes a multiagent learning framework augmented with a Collective Influence Estimation Network (CIEN). By explicitly modeling the collective influence of other agents on the task object, each agent can infer critical interaction information solely from its local observations and the task object's states, enabling efficient collaboration without explicit action information exchange. The proposed framework effectively avoids network expansion as the team size increases; moreover, new agents can be incorporated without modifying the network structures of existing agents, demonstrating strong scalability. Experimental results on multiagent cooperative tasks based on the Soft Actor-Critic (SAC) algorithm show that the proposed method achieves stable and efficient coordination under communication-limited environments. Furthermore, policies trained with collective influence modeling are deployed on a real robotic platform, where experimental results indicate significantly improved robustness and deployment feasibility, along with reduced dependence on communication infrastructure.

In domains such as biomedicine, materials, and finance, high-stakes deployment of large language models (LLMs) requires injecting private, domain-specific knowledge that is proprietary, fast-evolving, and under-represented in public pretraining. However, the two dominant paradigms for private knowledge injection each have pronounced drawbacks: fine-tuning is expensive to iterate, and continual updates risk catastrophic forgetting and general-capability regression; retrieval-augmented generation (RAG) keeps the base model intact but is brittle in specialized private corpora due to chunk-induced evidence fragmentation, retrieval drift, and long-context pressure that yields query-dependent prompt inflation. Inspired by how multimodal LLMs align heterogeneous modalities into a shared semantic space, we propose Generation-Augmented Generation (GAG), which treats private expertise as an additional expert modality and injects it via a compact, representation-level interface aligned to the frozen base model, avoiding prompt-time evidence serialization while enabling plug-and-play specialization and scalable multi-domain composition with reliable selective activation. Across two private scientific QA benchmarks (immunology adjuvant and catalytic materials) and mixed-domain evaluations, GAG improves specialist performance over strong RAG baselines by 15.34% and 14.86% on the two benchmarks, respectively, while maintaining performance on six open general benchmarks and enabling near-oracle selective activation for scalable multi-domain deployment.

Ruminal acidosis is a prevalent metabolic disorder in dairy cattle causing significant economic losses and animal welfare concerns. Current diagnostic methods rely on invasive pH measurement, limiting scalability for continuous monitoring. We present FUME (Fused Unified Multi-gas Emission Network), the first deep learning approach for rumen acidosis detection from dual-gas optical imaging under in vitro conditions. Our method leverages complementary carbon dioxide (CO2) and methane (CH4) emission patterns captured by infrared cameras to classify rumen health into Healthy, Transitional, and Acidotic states. FUME employs a lightweight dual-stream architecture with weight-shared encoders, modality-specific self-attention, and channel attention fusion, jointly optimizing gas plume segmentation and classification of dairy cattle health. We introduce the first dual-gas OGI dataset comprising 8,967 annotated frames across six pH levels with pixel-level segmentation masks. Experiments demonstrate that FUME achieves 80.99% mIoU and 98.82% classification accuracy while using only 1.28M parameters and 1.97G MACs--outperforming state-of-the-art methods in segmentation quality with 10x lower computational cost. Ablation studies reveal that CO2 provides the primary discriminative signal and dual-task learning is essential for optimal performance. Our work establishes the feasibility of gas emission-based livestock health monitoring, paving the way for practical, in vitro acidosis detection systems. Codes are available at https://github.com/taminulislam/fume.

Recently, self-play fine-tuning (SPIN) has been proposed to adapt large language models to downstream applications with scarce expert-annotated data, by iteratively generating synthetic responses from the model itself. However, SPIN is designed to optimize the current reward advantages of annotated responses over synthetic responses at hand, which may gradually vanish during iterations, leading to unstable optimization. Moreover, the utilization of reference policy induces a misalignment issue between the reward formulation for training and the metric for generation. To address these limitations, we propose a novel Triplet-based Self-Play fIne-tuNing (T-SPIN) method that integrates two key designs. First, beyond current advantages, T-SPIN additionally incorporates historical advantages between iteratively generated responses and proto-synthetic responses produced by the initial policy. Even if the current advantages diminish, historical advantages remain effective, stabilizing the overall optimization. Second, T-SPIN introduces the entropy constraint into the self-play framework, which is theoretically justified to support reference-free fine-tuning, eliminating the training-generation discrepancy. Empirical results on various tasks demonstrate not only the superior performance of T-SPIN over SPIN, but also its stable evolution during iterations. Remarkably, compared to supervised fine-tuning, T-SPIN achieves comparable or even better performance with only 25% samples, highlighting its effectiveness when faced with scarce annotated data.

The integration of large language models (LLMs) into autonomous agents has enabled complex tool use, yet in high-stakes domains, these systems must strictly adhere to regulatory standards beyond simple functional correctness. However, existing benchmarks often overlook implicit regulatory compliance, thus failing to evaluate whether LLMs can autonomously enforce mandatory safety constraints. To fill this gap, we introduce LogiSafetyGen, a framework that converts unstructured regulations into Linear Temporal Logic oracles and employs logic-guided fuzzing to synthesize valid, safety-critical traces. Building on this framework, we construct LogiSafetyBench, a benchmark comprising 240 human-verified tasks that require LLMs to generate Python programs that satisfy both functional objectives and latent compliance rules. Evaluations of 13 state-of-the-art (SOTA) LLMs reveal that larger models, despite achieving better functional correctness, frequently prioritize task completion over safety, which results in non-compliant behavior.

Optical spectroscopy based on second-order nonlinearity is a critical technique for characterizing two-dimensional (2D) crystals as well as bioimaging and quantum optics. It is generally believed that second-harmonic generation (SHG) in centrosymmetric crystals, such as graphene and other bilayer 2D crystals, is negligible without externally breaking the inversion symmetry. Here, we show that with a new homodyne detection technique, we can apparently circumvent this symmetry-imposed constraint and observe robust SHG in pristine centrosymmetric crystals, without any symmetry-breaking field. With its exceptional sensitivity, we resolve polarization-resolved SHG in bilayer hexagonal boron nitride (h-BN), bilayer 2H-WSe$_2$, and remarkably, Bernal-stacked bilayer graphene, allowing us to unambiguously identify the crystallographic orientation in these crystals via SHG for the first time. We also demonstrate that the new technique can be used to non-invasively detect uniaxial strain and optical geometric phase in these crystals. The observed SHG in our experiments is attributed to second-order nonlinearity in the quadrupole channel, which is controlled by the presence of the $C_2$ symmetry instead of the inversion symmetry. Our new technique expands the capability of nonlinear optical spectroscopy to encompass a large class of centrosymmetric materials that could never be measured before, and can be used for quantum sensing of moiré materials and twisted epitaxial films.

At scales larger than the forcing scale, some out-of-equilibrium turbulent systems (such as hydrodynamic turbulence, wave turbulence, and nonlinear optics) exhibit a state of statistical equilibrium where energy is equipartitioned among large-scale modes, in line with the Rayleigh-Jeans spectrum. Key open questions now pertain to either the emergence, decay, collapse, or other nonstationary evolutions from this state. Here, we experimentally investigate the free decay of large-scale hydroelastic turbulent waves, initially in a regime of statistical equilibrium. Using space- and time-resolved measurements, we show that the total energy of these large-scale tensional waves decays as a power law in time. We derive an energy decay law from the theoretical initial equilibrium spectrum and the linear viscous damping, as no net energy flux is carried. Our prediction then shows a good agreement with experimental data over nearly two decades in time, for various initial effective temperatures of the statistical equilibrium state. We further identify the dissipation mechanism and confirm it experimentally. Our approach could be applied to other decaying turbulence systems, with the large scales initially in statistical equilibrium.

Even-denominator fractional quantum Hall (FQH) states can be viewed as topological superconductors of composite fermions, supporting a charged chiral mode and $|\mathcal{C}_{cf}|$ neutral Majorana modes set by the Chern number $\mathcal{C}_{cf}$. Despite ongoing efforts, distinguishing the many competing paired phases remains an open problem. In this work, we propose a unified theory of charge transport across a quantum point contact (QPC) for general paired FQH states described by an $so(N)_1 \times u(1)$ conformal field theory. We derive the boundary effective action for an arbitrary number of Majorana fermions $N=|\mathcal{C}_{cf}|$ and develop a non-perturbative instanton approximation to describe tunneling processes. We establish a weak-strong duality relating strong quasiparticle tunneling to weak electron tunneling. We calculate the scaling dimensions of the tunneling operators and demonstrate that while the weak-coupling fixed point is generally unstable, the strong-coupling fixed point is stable for physically relevant filling fractions and number of Majorana fermions. These transport exponents provide a distinct experimental fingerprint to identify the topological phases of even-denominator FQH states.

Biased sampling in molecular dynamics simulations overcomes timescale limitations and delivers free-energy landscapes, essential to understand complex atomistic phenomena. However, when applied across diverse systems and processes, biasing protocols often require time- and resource-consuming fine-tuning. In search for robustness, we boost a prominent biasing method, Umbrella Sampling. To estimate the value of an integral, i.e., the free energy, our Bayesian Umbrella Quadrature (BUQ) method iteratively selects gradient samples, i.e., bias locations, that most reduce the posterior integral variance based on a noise-tolerant Gaussian process model, which also effectively interpolates between samples. We validate the method for a conformational change in a small peptide, a water-to-ice phase transition, and a substitution chemical reaction; obtaining excellent accuracies and speedups. To ease adoption of this more automated and universal free-energy method, we interface BUQ with wide-spread simulation packages and share hyperparametrization guidelines.

Magnetic phases are commonly identified through macroscopic magnetization, yet many ordered states, including antiferromagnets and altermagnets, possess a vanishing net moment despite distinct local spin structure. We show that such an order can be accessed through the measurement-induced steady state of a single primary qubit locally coupled to a spin lattice. Using a controlled primary-ancillary qubit protocol, we derive analytically that the steady state \emph{encodes} a locally weighted exchange field in a signed observable that is linear in the weak-coupling regime. Numerical simulations demonstrate lattice-scale resolution of antiferromagnetic and altermagnetic textures and robustness against short-correlated noise. Our results establish measurement-induced dissipation as a resource for detecting magnetic order through microscopic structure rather than through global moments.

The canonical model of stochastic search tracks a randomly diffusing "searcher" until it finds a "target." Owing to its many applications across science and engineering, this perennially popular problem has been thoroughly investigated in a variety of models. However, aside from some exactly solvable one-dimensional examples, very little is known if the searcher diffusivity varies in space. For such space-dependent or "heterogeneous" diffusion, one must specify the interpretation of the multiplicative noise, which is termed the Itô-Stratonovich dilemma. In this paper, we investigate how stochastic search with space-dependent diffusivity depends on this interpretation. We obtain general formulas for the probability distribution and all the moments of the stochastic search time and the so-called splitting probabilities assuming that the targets are small or weakly reactive. These asymptotic results are valid for general space-dependent diffusivities in general domains in any space dimension with targets of general shape which may be in the interior or on the boundary of the domain. We illustrate our theory with stochastic simulations. Our analysis predicts that stochastic search can depend strongly and counterintuitively on the multiplicative noise interpretation.

Spin transfer torques (STTs) control magnetisation by electric currents, enabling a range of nano-scale spintronic applications. They can destabilise the equilibrium magnetisation state by counteracting magnetic relaxation. Here, we maximise the STT effect through a dedicated growth-annealing protocol for CoFeB thin films, such that magnetic anisotropies originating from the interface and shape almost cancel each other. The nearly isotropic magnets enable low-current dynamical stabilisation of the magnetisation in the direction opposite to an applied magnetic field, thereby realising a spintronic analogue of the Kapitza pendulum. In an intermediate current regime, the STT drives large magnetisation vector fluctuations that cover the entire Bloch sphere. The continuous variable associated with the stochastic magnetisation direction may serve as a resource for probabilistic computing and neuromorphic hardware. Our results establish isotropic magnets as a platform to study as-yet-uncharted, far-from-equilibrium spin dynamics including anti-magnonics, with promising implications for unconventional computing paradigms.

Quantum devices require coherent coupling over macroscopic distances. Recently, resonances due to Klein tunneling and Andreev reflection states (KARS) have been observed in a naturally occurring p-n junction at the interface between Bi$_2$Sr$_2$CaCu$_2$O$_{8+x}$ (BSCCO), a high-Tc superconductor (HTS), and graphene. The resonances appear as conductance oscillations with gating. Here, we show coupling between the KARS in BSCCO-graphene-BSCCO devices of varying separation (L). The coupling is evidenced by a power-law decay of resonance period as L increases from tens of nanometers to single microns. These results demonstrate the long-distance coupling of KARS cavities in graphene-HTS junctions. The length dependence seen in experiments is supported by single-particle spectral functions which show KARS are coupled by transport modes in graphene. The strong coupling between KARS in BSCCO-graphene-BSCCO junctions showcases the novelty of HTS-graphene junctions for quantum circuits and unconventional Josephson junctions.

Double-helix (DH) nanowires provide a platform to study the influence of geometric chirality on spin chirality. Their three-dimensional (3D) helical architecture and tunable inter-strand coupling enable control of spin chirality, including the stabilization of topological 3D magnetic states such as helical domains and domain walls, topological stray fields, and extended helical vortex/skyrmion tubes. So far, the study of these and other 3D nanostructures is usually confined to a limited number of magnetic microscopy experiments in large facilities. Here, we investigate the reversal mechanism of a single DH nanowire using Dark-Field magneto-optical Kerr effect (DF-MOKE) magnetometry under external 3D magnetic fields. By analyzing the angular dependence of the DF-MOKE signal, we fit the reversal process using established models for domain-wall nucleation and propagation, finding a characteristic behavior similar to that reported for cylindrical nanowires. Micromagnetic simulations indicate that the reversal process goes through nucleation of the helical vortex tube in a curling manner while ptychographic X-ray magnetic circular dichroism data reveal that this helical vortex tube state forms through a mixed nucleation-propagation process. These observations provide a consistent microscopic picture of reversal mediated by a helical vortex tube extending along the nanowire. Our work provides a comprehensive characterization of magnetization reversal in DH nanowires and demonstrates that DF-MOKE magnetometry is effective for probing reversal mechanisms in single 3D nanostructures. This lab-based approach expands the range of accessible experiments beyond large-scale facilities, enabling extensive exploration of the rich spin states supported by 3D nano-geometries.

We study an open quantum spin chain with non-reciprocal dissipation using a theoretical approach known as time-dependent generalized Gibbs ensemble. In the regime of weak dissipation the system is fully characterized by its rapidity distribution and we derive a closed set of coupled differential equations governing their time evolution. We check the accuracy of this theory by benchmarking the results against numerical simulations. Using this framework we are able to compute both the magnetization density and current dynamics, identifying some relations between the two. The problem of the anomalous power-law exponents identified in a previous work is discussed. Our work constitutes a theoretical approach that is able to describe the physics of non-reciprocal open quantum spin chains beyond analyses based on non-interacting fermions.

This Ph.D. thesis reports on progress in rigorously establishing hydrodynamic principles from the microscopic Hamiltonian dynamics of quantum many-body systems in a general, non-model-specific manner. Using the C*-algebra framework of statistical mechanics, we treat systems directly in the thermodynamic limit, primarily focusing on quantum lattice models where tools such as Lieb-Robinson bounds yield rigorous statements. We thus provide a proof-of-principle that large-scale behaviours can indeed be seen as emerging from microscopic dynamics, with mathematical proof. We first report on ergodicity results in short-range models with exponentially decaying or finite-range interactions. We show that time-averaged observables converge to their ensemble averages and decorrelate from all other observables almost everywhere within the light-cone defined by Lieb-Robinson bounds. This relaxation property indicates the loss of information at large scales, from which we prove a Boltzmann-Gibbs principle: at the Euler scaling limit of large time and distance, observables project onto hydrodynamic modes (extensive conserved quantities), within correlation functions. These results hold independently of microscopic details, capturing the physical idea that such details are lost at large space-time scales. Regarding finer scales of hydrodynamics, we discuss rigorous lower bounds on the strength of diffusion. We establish a general result on the clustering of n-th order connected correlations within C* dynamical systems. These results are applied to obtain a strictly positive lower bound on the diffusion constant of chaotic open quantum spin chains with nearest-neighbor interactions. This thesis underlines the universality of hydrodynamic principles, provides a framework for establishing them rigorously, and sets the stage for future progress toward the goal of proving the hydrodynamic equations.

The complexity of biomolecular interactions necessitates advanced methodologies to accurately capture their behavior in solution. In this work, we focus on monoclonal antibodies and adopt a multi-scale coarse-graining strategy for their modeling, with particular emphasis on the role of electrostatic interactions. Using scattering experiments, theoretical analysis, and large-scale computer simulations, we explicitly compare two selected case studies-markedly different in their charge distributions. Through mutually corroborating lines of evidence, we demonstrate that conventional approaches relying on electrostatic screening and implicit charge representations fail to capture the structural and thermodynamic properties of antibody solutions when strong charge heterogeneity is present, even at a moderate (amino acid) level of coarse-graining. These findings highlight the importance of a correct treatment of electrostatic interactions and ion screening for heterogeneously- and oppositely-charged colloidal and protein systems. Such considerations are essential to move beyond descriptive models towards a truly predictive framework, with direct implications for the formulation of therapeutics and the treatment of other complex soft-matter systems.

Hour-glass clocks do not free-run; they keep time by riding an external rhythm. Motivated by the primordial KaiBC system in cyanobacteria, we study a driven, finite-state molecular clock that advances through a small number of biochemical states under an intrinsic driving energy and a rotating energy landscape set by day-night metabolism. In the continuum limit, coherence is maximized at a resonant operating point where the intrinsic drift matches the driving frequency. In realistic clocks with a finite number of states, discreteness matters: as the rotating landscape sweeps over a lattice of states, it generates a small and high frequency vibration of the collective phase that makes timing inaccurate. Combining the resonant cost with this discreteness penalty yields a trade-off in the number of states: few states are energetically cheap but noisy; many states are precise but costly. The optimum lies at moderate discretization (typically five to fifteen states) and an environmental coupling that is strong enough for responsiveness yet weak enough to avoid large discrete-state vibrations. These design rules rationalize why KaiC's hexameric architecture falls near the predicted optimum and suggest a general principle for hour-glass clocks across organisms.

When we consider canonical averages for classical discrete systems, typically referred to as substitutional alloys, the map from many-body interatomic interactions to thermodynamic equilibrium configurations generally exhibits complicated nonlinearity. This canonical nonlinearity is fundamentally rooted in deviations of the discrete configurational density of states (CDOS) from continuous Gaussian families, and has conventionally been characterized by the Kullback-Leibler (KL) divergence on discrete statistical manifold. Thus, the previous works inevitablly missed intrinsic nonlinearities induced by discretization of Gaussian families, which remains invisible within conventional information-geometric descriptions. In the present work, we identify and quantify such unavoidable canonical nonlinearity by employing the 2-Wasserstein distance with a cost function aligned with the Fisher metric for Gaussian families. We derive an explicit expression for the Wasserstein distance in the limit of vanishing discretization scale d to 0: W2 = d*sqrt(Tr(Gamma)^(-1)/12), where Gamma denotes covariance matrix of the CDOS. We further show that this limiting Wasserstein distance admits a clear geometric interpretation on the statistical manifold, equivalent to a KL divergence associated with the expected parallel translations of continuous Gaussian. Our framework thus provides a transport-information-geometric characterization of discretization-induced nonlinearity in classical discrete systems, with future potential applications to configurational thermodynamics.

It was conjectured for a long time that the tectonic plates are in a self-organized state of criticality and that the Gutenberg-Richter (power) law is a manifestation of that. It was recently shown that for a system near criticality, the inequality of their responses toward external driving could indicate proximity to the critical point. In this work, we show with numerical simulations and seismic data analysis that large earthquake events have a tendency to follow events that are highly unequal. We have applied this framework to various tectonically active regions, such as North America, Southern Japan, parts of South-East Asia and Indonesia.

The eigenstate thermalization hypothesis (ETH) posits how isolated quantum many-body systems thermalize, assuming that individual eigenstates at the same energy density have identical expectation values of local observables in the limit of large systems. While the ETH apparently holds across a wide range of interacting quantum systems, in this work we show that it requires generalization in the presence of thermal first-order phase transitions. We introduce a class of all-to-all spin models, featuring first-order thermal phase transitions that stem from two distinct mean-field solutions (two ``branches'') that exchange dominance in the many-body density of states as the energy is varied. We argue that for energies in the vicinity of the thermal phase transition, eigenstate expectation values do not need to converge to the same thermal value. The system has a regime with coexistence of two classes of eigenstates corresponding to the two branches with distinct expectation values at the same energy density, and another regime with Schrodinger-cat-like eigenstates that are inter-branch superpositions; these two regimes are separated by an eigenstate phase transition. We support our results by semiclassical calculations and an exact diagonalization study of a microscopic spin model, and argue that the structure of eigenstates in the vicinity of thermal first-order phase transitions can be experimentally probed via non-equilibrium dynamics.

Defect engineering via parallel cracks has been proposed as a route to tailor the fracture response of graphene. However, atomistic fracture predictions can be strongly sensitive to the interatomic potential. Here, we quantify the effect of potential choice by revisiting H-passivated graphene containing two parallel cracks separated by a gap $W_{\text{gap}}$ loaded in tension along the armchair (AC) and zigzag (ZZ) directions. Molecular dynamics simulations using the AIREBO potential under the same geometry and loading protocol previously studied with ReaxFF, are employed, so enabling a direct comparison. Stress-strain responses, Young's modulus, an effective mode-I stress intensity factor, and energy absorption are evaluated as functions of $W_{\text{gap}}$. Compared with ReaxFF, AIREBO predicts lower peak stresses and earlier catastrophic softening, leading to reduced post-peak deformation capacity and energy absorption. Ductility and energy absorption are shown to be highly potential-dependent, underscoring the need for careful potential selection in defect-engineered graphene fracture simulations.

We study the spectral statistics and wave-function properties of a one-dimensional quantum system subject to a Cantor-type fractal potential. By analyzing the nearest-neighbor level spacings, inverse participation ratio (IPR), and the scaling behavior of the integrated density of states (IDS), we demonstrate how the self-similar geometry of the potential is imprinted on the quantum spectrum. The energy-resolved level spacings form a hierarchical, filamentary structure, in sharp contrast to those of periodic and random systems. The normalized level-spacing distribution exhibits a bimodal structure, reflecting the deterministic recurrence of spectral gaps. A multifractal analysis of eigenstates reveals critical behavior: the generalized fractal dimensions $D_q$ lie strictly between the limits of extended and localized states, exhibiting a distinct $q$-dependence. Consistently, the IPR indicates the coexistence of quasi-extended and localized features, characteristic of critical wave functions. The IDS shows anomalous power-law scaling at low energies, with an exponent close to the Hausdorff dimension of the underlying Cantor set, indicating that the geometric fractality governs the spectral dimensionality. At higher energies, this scaling crosses over to the semiclassical Weyl law. Our results establish a direct connection between deterministic fractal geometry, hierarchical spectral statistics, and quantum criticality.

Filamentous bacteriophages fd are viral particles, highly monodisperse in size, that have been widely used as a model colloidal system for studying the self-assembly of rod-shaped particles as well as a versatile template in nanoscience. In aqueous suspensions, fd viruses exhibit lyotropic behavior, forming liquid crystalline phases as their concentration increases. Here, we report a solvent-free system displaying thermotropic phase behavior, achieved through covalent coupling of low molecular weight PEG-based polymer surfactant onto the fd virus surface. Upon lyophilization of aqueous suspensions of these polymer-grafted bacteriophages and subsequent thermal annealing, a solvent-free material is obtained, exhibiting both viscoelasticity and, notably, thermotropic liquid crystalline properties. A combination of small-angle X-ray scattering and optical microscopy experiments reveals the formation of an ordered hexagonal mesophase below 30 °C, which undergoes a melting transition into an isotropic liquid at higher temperatures. Our results demonstrate an efficient approach for converting lyotropic into thermotropic phase behavior in the columnar liquid crystalline phase of filamentous fd colloids. This approach paves the way for extending such functionalization to other technologically relevant rod-like systems, such as carbon nanotubes and cellulose nanocrystals, enabling the introduction of thermotropic properties in anhydrous colloidal materials.

The anomalous Hall effect has been understood in terms of the geometric nature of Bloch bands and impurity scattering, and has been observed in a wide variety of magnetic materials such as ferromagnets and antiferromagnets. Recently, a large anomalous Hall effect was reported in the noncollinear antiferromagnetic metal NbMnP whose magnetic order is a mixture of the even-parity and the odd-parity magnetic components. Such a magnetic structure is expected to exhibit the anomalous Hall effect and the nonlinear Hall effect from the symmetry breaking of the antiferromagnet ordering. Here, we theoretically investigate the intrinsic anomalous and nonlinear Hall effect of NbMnP induced by the quantum geometry of Bloch band using the first-principles calculation and the Wannier interpolation method. We found that the intrinsic Hall response of NbMnP is predominantly governed by the strongly enhanced Berry curvature and Berry-connection-polarization dipole on a specific mirror plane. These enhanced geometric quantities originate from the spin-orbit-coupling-induced gap openings along the nodal lines. Our results indicate that NbMnP serves as a model system for investigating transport phenomena originating from nodal-lines in parity-mixed antiferromagnets.

Ion transport underlies the operation of biological ion channels and governs the performance of electrochemical energy-storage devices. A long-standing anomaly is that smaller alkali metal ions, such as Li$^+$, migrate more slowly in water than larger ions, in apparent violation of the Stokes-Einstein relation. This breakdown is conventionally attributed to dielectric friction, a collective drag force arising from electrostatic interactions between a drifting ion and its surrounding solvent. Here, combining nanopore transport measurements over electric fields spanning several orders of magnitude with molecular dynamics simulations, we show that the time-averaged electrostatic force on a migrating ion is not a drag force but a net driving force. By contrasting charged ions with neutral particles, we reveal that ionic charge introduces additional Lorentzian peaks in the frequency-dependent friction coefficient. These peaks originate predominantly from short-range Lennard-Jones (LJ) interactions within the first hydration layer and represent additional channels for energy dissipation, strongest for Li$^+$ and progressively weaker for Na$^+$ and K$^+$. Our results demonstrate that electrostatic interactions primarily act to tighten the local hydration structure, thereby amplifying short-range LJ interactions rather than directly opposing ion motion. This microscopic mechanism provides a unified physical explanation for the breakdown of the Stokes-Einstein relation in aqueous ion transport.

Remarkably, even under negligible inertia, the addition of microstructural agents can generate chaotic flow fields. Such behavior can arise in polymer solutions, leading to elastic turbulence, or from active, self-driven particles, which generate active turbulence. Here, we demonstrate a close and hitherto unrecognized connection between these two classes of turbulence. Specifically, we reveal that their continuum descriptions are analogous at the macroscopic level, such that polymeric fluids can be interpreted as a deformable analogue of contractile active matter. Moreover, our numerical results for Kolmogorov flow demonstrate that the transition into the well-known traveling arrowhead structures in elastic turbulence is marked by the emergence of $\pm 1/2$ topological defects, long recognized as a defining feature of active turbulence, in the polymer director field. Importantly, these coherent structures originate from a transverse instability driven by activity-like gradients generated by anisotropically stretched, contractile polymers. At sufficiently strong activity, the system undergoes a transition into a flow-suppressed state characterized by weak polymer stretching and ordering, a behavior that can be explained by analogy with the spontaneous-flow transition observed in channel-confined active nematics.

Bell or Clauser-Horne-Shimony-Holt (CHSH) tests on superconducting quantum processors are commonly interpreted under the assumption that repeated circuit executions sample a single, stationary preparation ensemble. Here we show that this assumption can be violated on contemporary hardware, with direct implications for the interpretation of observed Bell violations. We introduce an ensemble-divergence framework in which slow temporal drift of the preparation process induces context-dependent effective ensembles, even when measurement independence and locality are preserved. This leads to a relaxed Bell bound $|S| \le 2 + 6δ_{\mathrm{ens}}$, where $δ_{\mathrm{ens}}$ quantifies preparation nonstationarity. Because $δ_{\mathrm{ens}}$ is not directly observable, we develop an operational witness $δ_{\mathrm{op}}$ based on bin-resolved outcome statistics for fixed measurement channels. Using Pauli-axis measurements on IBM superconducting processors, we observe statistically significant operational drift that persists after full two-qubit readout mitigation, ruling out measurement artifacts. In contrast, drift extracted from CHSH-optimal measurements is eliminated by mitigation, demonstrating that such settings are unsuitable for diagnosing preparation nonstationarity. We further show that the observed Bell violations imply only modest ensemble divergences, comparable in scale to those required in Hall-type measurement-dependence models, but arising here solely from preparation drift combined with experimental scheduling. Our results identify a preparation-dependent loophole relevant to Bell tests on noisy intermediate-scale quantum devices and highlight the necessity of drift-aware protocols for reliable quantum certification.

Topology provides a unifying framework for understanding robust transport through protected edge states arising from nontrivial wavenumber topology. Extending these concepts to active matter, however, remains largely unexplored experimentally, with realizations limited to systems composed of chiral active particles. Here, we realize topological edge states in dense bacterial suspension, which represents a prototypical active matter system, using microfabricated geometrical structures with nontrivial wavenumber topology. Inspired by previous theoretical studies, we constructed a directional kagome network composed of ratchet-shaped channels that induce unidirectional bacterial flow. In this network, we found clear edge localization of bacterial density. A steady-state analysis based on the bacterial transport model and experimentally measured velocity field reveals how the characteristic collective flow generates edge localization. The model also uncovers the topological origin of the observed edge states. By tuning the geometry of the microfabricated networks, we identified directional channel design and network chirality as the key design features essential for the emergence of the edge state. Our results pave the way for establishing a control and design principle of topological transport in such active matter systems.

We demonstrate a dissipative protocol for ground-state preparation of a quantum spin chain on a trapped-ion quantum computer. As a first step, we derive a Kraus representation of a dissipation channel for the protocol recently proposed by Ding et al. [Phys. Rev. Res. 6, 033147 (2024)] that still holds for arbitrary temporal discretization steps, extending the analysis beyond the Lindblad dynamics regime. The protocol guarantees that the fidelity with the ground state monotonically increases (or remains unchanged) under repeated applications of the channel to an arbitrary initial state, provided that the ground state is the unique steady state of the dissipation channel. Using this framework, we implement dissipative ground-state preparation of a transverse-field Ising chain for up to 19 spins on the trapped-ion quantum computer Reimei provided by Quantinuum. Despite the presence of hardware noise, the dynamics consistently converges to a low-energy state far away from the maximally mixed state even when the corresponding quantum circuits contain as many as 4110 entangling gates, demonstrating the intrinsic robustness of the protocol. By applying zero-noise extrapolation, the resulting energy expectation values are systematically improved to agree with noiseless simulations within statistical uncertainties.

We study the Brownian motion of a rigid rod threading through a small fixed ring while the ring can freely rotate. We derive the distribution function for the sliding displacement and the unit vector along the rod both at equilibrium and non-equilibrium. The equilibrium distribution is quadratic in the sliding displacement and is controlled by the moment of inertia (mass distribution). Applying the Onsager variational principle, we derive a Smoluchowski equation in which sliding and rotational diffusion are coupled. The mean square displacement (MSD) of sliding shows a metastable plateau in a certain time range before it approaches the final equilibrium value. The longest sliding relaxation time decreases as $α^{-1/2}$ as the moment of inertia increases. The rotational relaxation time obtained from the orientational correlation function is longer than that of a rod with its center fixed but faster than a rod with one end fixed. These results may be useful in understanding the dynamics of polymers connected by sliding rings.

Energy-efficient fast switching of spin orientations or textures is a core issue of spintronics, which is highly demanded but remains challenging. Different from the mainstream routes based on spin-transfer torque or spin-orbit torque, here we propose another mechanism coined as magnetoelectric torque to switch the magnetization in polar magnetic bilayers via pure electric field. In some magnetic van der Waals bilayers, when the electrostatic energy of polarization can compensate the interlayer magnetic coupling, a magnetoelectric torque is generated to fastly flip spins within a few picoseconds, which is demonstrated by combining the first-principles calculations, analytic model, as well as atomistic simulations. Such a magnetoelectric torque doesn't rely on the spin-orbit coupling and is generally active in polar magnetic homostructures and heterostructures. Our work provides an alternative route to switch magnetization in nanoscale, which may benefit the energy-saving and fast-response spintronic devices.

When a liquid droplet completely wets a hard substrate, its spreading dynamics follow Tanner's law, with the droplet radius growing as the one-tenth power of time. Here, we investigate how these dynamics change when silicone oil droplets spread on soft silicone elastomer and gel films supported by a rigid silicon substrate. While the droplets fully wet the elastomer surface, they also simultaneously swell the elastomer film. By varying the film thickness, we observe deviations from the classical power-law scaling, which we interpret in terms of changes to the effective stiffness and the absorption potential of the system. We describe the spreading behavior using a phenomenological model that accounts for both absorption and mechanical contributions.

Discrete charge fluctuations, routinely observed in semiconductor quantum dot devices, may contribute significantly to device drift and errors resulting from qubit miscalibration. Understanding the nature and origins of these discrete charge fluctuations may provide insights into material improvements or means of mitigating charge noise in semiconductor quantum dot devices. In this work, we measure multi-level charge fluctuations present in a Si/SiGe double quantum dot device over a range of device operating voltages and temperatures. To characterize the parameter-dependent dynamics of the underlying fluctuating degrees of freedom, we perform a detailed analysis of the measured noise timeseries. We perform algorithmically assisted drift detection and change point detection to detrend the data and remove a slow fluctuator component, as a preprocessing step. We perform model comparison on the post-processed time series between different $n$-level fluctuator ($n$LF) factorial hidden Markov models (FHMMs), finding that although at most sweep values the independent pair of 2LFs model would be preferred, in a particular region of voltage space the 4LF model outperforms the other models, indicating a conditional rate dependence between the two fluctuators. By tracking fluctuator transition rates, biases, and weights over a range of different device configurations, we estimate gate voltage and conductivity sensitivity. In particular, we fit a phenomenological, detailed balance model to the extracted independent 2LFs rate data, yielding lever arm estimates in the range of $-2 μ$eV/mV up to $4 μ$eV/mV between the two 2LFs and nearby gate electrodes. We expect that these characterization results may aid in subsequent spatial triangulation of the charge fluctuators.

The accurate identification and control of spatiotemporal chaos in reaction-diffusion systems remains a grand challenge in chemical engineering, particularly when the underlying catalytic surface possesses complex, unknown topography. In the \textit{Defect Turbulence} regime, system dynamics are governed by topological phase singularities (spiral waves) whose motion couples to manifold curvature via geometric pinning. Conventional Physics-Informed Neural Networks (PINNs) using ReLU or Tanh activations suffer from fundamental \textit{spectral bias}, failing to resolve high-frequency gradients and causing amplitude collapse or phase drift. We propose a Multi-Scale SIREN-PINN architecture leveraging periodic sinusoidal activations with frequency-diverse initialization, embedding the appropriate inductive bias for wave-like physics directly into the network structure. This enables simultaneous resolution of macroscopic wave envelopes and microscopic defect cores. Validated on the complex Ginzburg-Landau equation evolving on latent Riemannian manifolds, our architecture achieves relative state prediction error $ε_{L_2} \approx 1.92 \times 10^{-2}$, outperforming standard baselines by an order of magnitude while preserving topological invariants ($|ΔN_{defects}| < 1$). We solve the ill-posed \textit{inverse pinning problem}, reconstructing hidden Gaussian curvature fields solely from partial observations of chaotic wave dynamics (Pearson correlation $ρ= 0.965$). Training dynamics reveal a distinctive Spectral Phase Transition at epoch $\sim 2,100$, where cooperative minimization of physics and geometry losses drives the solver to Pareto-optimal solutions. This work establishes a new paradigm for Geometric Catalyst Design, offering a mesh-free, data-driven tool for identifying surface heterogeneity and engineering passive control strategies in turbulent chemical reactors.

Urban scaling theory posits that urban indicators follow power-law relations with population, yet the evolution of these patterns - and the role of regional differences in settings marked by social inequalities and unplanned urbanization - remains poorly understood. Here, we analyze nearly three decades of mortality data from Brazilian cities to investigate the scaling of external causes of death: homicides, suicides, and accidents. Using a hierarchical Bayesian framework and spatial correlation analysis, we find that these mortality indicators exhibit distinct, regionally heterogeneous scaling trajectories. Homicide mortality has significantly attenuated its typical superlinear scaling with increased spatial clustering, suggesting a redistribution of violence to smaller cities and intensified intercity interactions, possibly linked to the consolidation of organized crime. Suicide mortality, usually sublinear, has trended upward, implying a weakening of urban agglomerations' protective effect. Accident mortality remains superlinear, with transport fatalities scaling nearly proportionally, and non-transport accidents becoming superlinear. The scaling changes for suicides and accidents coincide with less correlated and stable spatial patterns, suggesting that the underlying processes predominantly operate within city boundaries. Finally, while scaling exponents have evolved more homogeneously across Brazilian states, scale-adjusted mortality remains highly heterogeneous, indicating that fundamental processes govern scaling laws, whereas state-specific factors drive scale-adjusted metrics.

We present a first-principles framework for the calculation of phonons in nanostructures with cyclic and/or helical symmetry. In particular, we derive a cyclic- and helical-symmetry-adapted representation of the dynamical matrix at arbitrary phonon wavevectors within a variationally formulated, symmetry-adapted density functional perturbation theory framework. In so doing, we also derive the acoustic sum rules for cylindrical geometries, which include a rigid-body rotational mode in addition to the three translational modes. We implement the cyclic- and helical-symmetry-adapted formalism within a high-order finite-difference discretization. Using carbon nanotubes as representative systems, we demonstrate the accuracy of the framework through excellent agreement with periodic plane-wave results. We further apply the framework to compute the Young's and shear moduli of carbon nanotubes, as well as the scaling laws governing the dependence of ring and radial breathing mode phonon frequencies on nanotube diameter. The elastic moduli are found to be in agreement with previous density functional theory and experimental results, while the phonon scaling laws show qualitative agreement with previous atomistic simulations.

Political corruption is inherently an affiliation process linking agents to corruption cases; yet it is often studied via one-mode projections that connect co-offenders within the same scandal, implying a loss of information that potentially confounds properties of agents and cases. Here, we adopt a bipartite representation to analyze datasets of corruption scandals in Brazil and Spain spanning nearly three decades. By tracking the temporal growth of these networks, we quantify density and redundancy measures to capture partner reuse and co-occurrence across cases. Networks in both countries become progressively sparser over time, and agent redundancy is systematically higher than case redundancy, indicating a small cadre of recidivists who recombine largely with novice partners rather than forming durable co-offending ties. These networks exhibit near-exponential degree distributions, reflecting low recidivism and likely high coordination costs and secrecy constraints of large-scale scandals. Our bipartite view further reveals a moderate cross-mode disassortative degree mixing between agents and cases, with high-degree agents distributing their activity across small cases and large scandals mainly comprising low-degree participants. Finally, identifying atypical individuals within the bipartite structure reveals criminal trajectories marked by a gradual rise in network embeddedness that can appear ordinary in agent-projected networks.

Excitons in monolayer transition-metal dichalcogenides (TMDs) have garnered significant attention because of their large binding energies due to weakly screened Coulomb interaction, and direct bandgap at the K/K$^\prime$ point in the hexagonal Brillouin zone featuring spin-polarised bands due to spin-orbit coupling and lack of inversion symmetry. This makes them prospective for next-generation optoelectronic and quantum devices. However, despite the intense research activity, the reported values for exciton binding energies, quasiparticle gaps, and spectral features exhibit substantial variation across both experimental and theoretical studies. In this article, we present a comprehensive and critical assessment of the current understanding of excitonic properties in single-layer TMDs, integrating results from the angle-resolved photoemission spectroscopy (ARPES), photoluminescence (PL) measurements, and other experimental techniques with first-principles theoretical insights. Special emphasis is placed on the comparison and reconciliation of discrepancies observed across different experimental setups and sample qualities. Furthermore, we highlight our state-of-the-art GW-BSE calculations, which include both equilibrium and laterally strained systems, to systematically analyse the behaviour of direct and indirect excitons. By evaluating the effect of strain as a tunable control variable, we demonstrate its potential to engineer excitonic properties, supported by cross-validation against prior theoretical predictions and experimental findings. In doing so, we clarify the sources of discrepancies in the literature and offer a unified perspective on excited-state engineering strategies in two-dimensional TMDs.

This work demonstrates efficient Forster resonance energy transfer (FRET) between ensembles of shallow nitrogen-vacancy (NV) centers located 7 nm and 9 nm below a single-crystal diamond surface and a naturally occurring fluorophore, namely a mixture of chlorophyll a and b molecules extracted from Arabidopsis thaliana. The broad fluorescence band of NV centers spectrally overlaps with the absorption of chlorophyll molecules, enabling FRET. As a result, depositing a chlorophyll layer on the diamond surface reduces the NV fluorescence lifetime from approximately 14 ns to below 4 ns, indicating efficient energy transfer. Laser-induced photobleaching of chlorophyll restores the unquenched NV lifetime. In contrast, NV centers located deeper within the diamond at depths of 40 nm and 72 nm remain unaffected, confirming that the observed quenching originates from a short-range FRET mechanism. The NV ensembles retain their optically detected magnetic resonance contrast while FRET is observed, demonstrating preservation of their spin properties. These proof-of-principle experiments establish the feasibility of combining FRET-based distance measurements with magnetic sensing using optically readable spins.

RNA polymerase (RNAP) is a molecular machine that reads information encoded in the base sequence of a DNA template while producing mechanical motion (transcription elongation; forward/backward stepping; backtracking) through chemical-potential differences of nucleoside triphosphates (NTPs) and fluctuations under external conditions. A prior work formulated a mismatch in free-energy accounting as the involvement of a term originating from genetic information (e.g.\ $k_BT\log P(N)$), and interpreted RNAP as a Maxwell's demon / Szilard-engine-like device that converts information into motion. However, in information thermodynamics, the bookkeeping of information and dissipation can depend on how one partitions the composite system into a device and a target (observer/observed labeling).

Accurate de novo molecular and materials design requires structure-property models that generalize beyond known regimes. Although pretrained atomistic models achieve strong in-distribution accuracy after fine-tuning, their reliability under out-of-distribution (OOD) conditions remains unclear. We identify a critical failure mode in downstream adaptation: standard fine-tuning induces representation collapse, erasing pretrained chemical and structural priors and severely degrading OOD performance. To address this limitation, we propose multi-task fine-tuning (MFT), which jointly optimizes downstream property prediction with a physically grounded force-field objective inherited from pretraining. This approach preserves essential chemical priors while enabling task-specific adaptation. Across molecular and materials benchmarks, MFT consistently improves OOD generalization, approaching the theoretical limit set by in-distribution accuracy, while outperforming standard fine-tuning, training from scratch, and state-of-the-art task-specific models. These results establish safe adaptation as a central requirement for large atomistic models and position MFT as a practical and data-efficient pathway toward robust molecular and materials discovery.

The Geant4 toolkit is the leading software for the simulation of particle transport through matter, widely used in nuclear physics, high-energy physics, and medical physics. However, the initial learning curve for new developers can be steep, often due to the complexity and experiment-specific nature of many introductory examples. This paper introduces MinGLE (Mini Geant4 Learning Example), a dedicated application template designed to be a universal, flexible, and educational starting point for Geant4 projects. MinGLE achieves a complete, functional simulation kernel using fewer than 70 lines of core C++ code. This minimalism is realized by leveraging contemporary Geant4 features, including factory classes for run management and physics, and the Text Geometry format for detector definition. Furthermore, MinGLE employs a unique pedagogical structure, using Git tags to document the incremental development of eleven core Geant4 components, with each tagged version being fully compilable, executable, and testable, providing a clear, step-by-step learning resource.

Accurate identification of chemical species and reaction pathways from molecular dynamics (MD) trajectories is a prerequisite for deriving predictive chemical-kinetic models and for mechanistic discovery in reactive systems. However, state-of-the-art trajectory analysis methods infer bonding from instantaneous distance thresholds, which can misclassify transient, nonreactive encounters as bonds and thereby introduce spurious intermediates, distorted reaction networks, and biased rate estimates. Here, we introduce ChemXDyn, a dynamics-aware computational methodology that leverages time-resolved interatomic distance signatures as a core principle to robustly identify chemically consistent bonded interactions and, consequently, extract meaningful reaction pathways. In particular, ChemXDyn propagates molecular connectivity through time while enforcing atomic valence and coordination constraints to distinguish genuine bond-breaking and bond-forming events from transient, nonreactive encounters. We evaluate ChemXDyn on ReaxFF MD simulations of hydrogen and ammonia oxidation and on neural-network potential MD simulations of methane oxidation, and benchmark its performance against widely used trajectory analysis methods. Across these cases, ChemXDyn suppresses unphysical species prevalent in static analyses, recovers experimentally consistent reaction pathways, and improves the fidelity of rate constant estimation. In ammonia oxidation, ChemXDyn removes unphysical intermediates and resolves key NOx- and N2O-forming and -consuming routes. In methane oxidation, it reconstructs the canonical progression from CH4 to CO2. By linking atomistic dynamics to chemically consistent reaction identification, ChemXDyn provides a transferable foundation for MD-derived reaction networks and kinetics, with potential utility spanning combustion, catalysis, plasma chemistry, and electrochemical environments.

Fluid-structure interaction (FSI) simulation of biological systems presents significant computational challenges, particularly for applications involving large structural deformations and contact mechanics, such as heart valve dynamics. Traditional ALE methods encounter fundamental difficulties with such problems due to mesh distortion, motivating immersed techniques. This work presents a novel open-source immersed FSI framework that strategically couples two mature finite element libraries: MFEM, a GPU-ready and scalable library with state-of-the-art parallel performance developed at Lawrence Livermore National Laboratory, and FEBio, a nonlinear finite element solver with sophisticated solid mechanics capabilities designed for biomechanics applications developed at the University of Utah. This coupling creates a unique synergy wherein the fluid solver leverages MFEM's distributed-memory parallelization and pathway to GPU acceleration, while the immersed solid exploits FEBio's comprehensive suite of hyperelastic and viscoelastic constitutive models and advanced solid mechanics modeling targeted for biomechanics applications. FSI coupling is achieved using a fictitious domain methodology with variational multiscale stabilization for enhanced accuracy on under-resolved grids expected with unfitted meshes used in immersed FSI. A fully implicit, monolithic scheme provides robust coupling for strongly coupled FSI characteristic of cardiovascular applications. The framework's modular architecture facilitates straightforward extension to additional physics and element technologies. Several test problems are considered to demonstrate the capabilities of the proposed framework, including a 3D semilunar heart valve simulation. This platform addresses a critical need for open-source immersed FSI software combining advanced biomechanics modeling with high-performance computing infrastructure.

Non-Newtonian fluids encompass a large family of fluids with additional nonlinear material properties, contributing to non-trivial flow behaviour that cannot be captured through a single constant viscosity term. Common non-Newtonian characteristics include shear-thinning, shear-thickening, viscoplasticity, and viscoelasticity, commonly encountered in everyday fluids, such as ketchup, blood, toothpaste, mud, etc., as well as practical applications involving porous media, cosmetics, food processing, and pharmaceuticals. Due to the complex nature of these fluids, accurate computational fluid dynamics simulations are essential for predicting their behaviour under various flow conditions. Recent advancements have highlighted the growing trend of using the lattice Boltzmann method to solve such complex flows, owing to its ability to handle intricate boundary conditions, ease of including additional multiphysics, and providing computationally efficient parallel simulations. Since the initial review over a decade ago [Phillips & Roberts, IMA J. Appl. Math. 76, 790-816 (2011)], significant advancements have been made to the lattice Boltzmann method to simulate non-Newtonian fluids. Here, we present a comprehensive review of different lattice Boltzmann techniques used to solve non-Newtonian fluid systems, specifically dealing with shear-dependent viscosity, viscoplasticity, and viscoelasticity. In addition, we discuss various benchmark cases that validate these approaches and highlight their growing application to realistic and challenging complex flow problems. We further address outstanding issues in current lattice Boltzmann models, as well as future directions for numerical advancement and application.

Autonomous experimentation holds the potential to accelerate materials development by combining artificial intelligence (AI) with modular robotic platforms to explore extensive combinatorial chemical and processing spaces. Such self-driving laboratories can not only increase the throughput of repetitive experiments, but also incorporate human domain expertise to drive the search towards user-defined objectives, including improved materials performance metrics. We present an autonomous materials synthesis extension to SARA, the Scientific Autonomous Reasoning Agent, utilizing phase information provided by an automated probabilistic phase labeling algorithm to expedite the search for targeted phase regions. By incorporating human input into an expanded SARA-H (SARA with human-in-the-loop) framework, we enhance the efficiency of the underlying reasoning process. Using synthetic benchmarks, we demonstrate the efficiency of our AI implementation and show that the human input can contribute to significant improvement in sampling efficiency. We conduct experimental active learning campaigns using robotic processing of thin-film samples of several oxide material systems, including Bi$_2$O$_3$, SnO$_x$, and Bi-Ti-O, using lateral-gradient laser spike annealing to synthesize and kinetically trap metastable phases. We showcase the utility of human-in-the-loop autonomous experimentation for the Bi-Ti-O system, where we identify extensive processing domains that stabilize $δ$-Bi$_2$O$_3$ and Bi$_2$Ti$_2$O$_7$, explore dwell-dependent ternary oxide phase behavior, and provide evidence confirming predictions that cationic substitutional doping of TiO$_2$ with Bi inhibits the unfavorable transformation of the metastable anatase to the ground-state rutile phase. The autonomous methods we have developed enable the discovery of new materials and new understanding of materials synthesis and properties.

Our brain functions as a complex communication network, and studying it from a network perspective offers valuable insights into its organizational principles and links to cognitive functions and brain disorders. However, most current network studies typically use brain regions as nodes, often overlooking the intricate folding patterns of finer-scale anatomical landmarks within these regions. In this study, we introduce a novel approach to integrate the brain's two primary folding patterns - gyri and sulci - into a unified network termed the Gyral-Sulcal-Net (GS-Net), in which three different types of finer-scale landmarks have been successfully identified. We evaluated the proposed GS-Net across multiple datasets, comprising over 1,600 brains, spanning different age groups (from 34 gestational weeks to elderly adults) and cohorts (healthy brains and those with pathological conditions). The experimental results demonstrate that the GS-Net can effectively integrate and represent diverse cortical folding patterns from a network perspective. More importantly, this approach offers a promising way for integrating different folding patterns into a unified anatomical brain network, alongside structural and functional networks, providing a comprehensive framework for studying brain networks. (The GS-Net toolbox will be released soon.)

Accurate individual identification is essential for monitoring rare amphibians, yet invasive marking is often unsuitable for critically endangered species. We evaluate state-of-the-art computer-vision methods for photographic re-identification of the Hula painted frog (Latonia nigriventer) using 1,233 ventral images from 191 individuals collected during 2013-2020 capture-recapture surveys. We compare deep local-feature matching in a zero-shot setting with deep global-feature embedding models. The local-feature pipeline achieves 98% top-1 closed-set identification accuracy, outperforming all global-feature models; fine-tuning improves the best global-feature model to 60% top-1 (91% top-10) but remains below local matching. To combine scalability with accuracy, we implement a two-stage workflow in which a fine-tuned global-feature model retrieves a short candidate list that is re-ranked by local-feature matching, reducing end-to-end runtime from 6.5-7.8 hours to ~38 minutes while maintaining ~96% top-1 closed-set accuracy on the labeled dataset. Separation of match scores between same- and different-individual pairs supports thresholding for open-set identification, enabling practical handling of novel individuals. We deploy this pipeline as a web application for routine field use, providing rapid, standardized, non-invasive identification to support conservation monitoring and capture-recapture analyses. Overall, in this species, zero-shot deep local-feature matching outperformed global-feature embedding and provides a strong default for photo-identification.

Accurate and generalisable segmentation of stroke lesions from magnetic resonance imaging (MRI) is essential for advancing clinical research, prognostic modelling, and personalised interventions. Although deep learning has improved automated lesion delineation, many existing models are optimised for narrow imaging contexts and generalise poorly to independent datasets, modalities, and stroke stages. Here, we systematically evaluated stroke lesion segmentation using the nnU-Net framework across multiple heterogeneous, publicly available MRI datasets spanning acute and chronic stroke. Models were trained and tested on diffusion-weighted imaging (DWI), fluid-attenuated inversion recovery (FLAIR), and T1-weighted MRI, and evaluated on independent datasets. Across stroke stages, models showed robust generalisation, with segmentation accuracy approaching reported inter-rater reliability. Performance varied with imaging modality and training data characteristics. In acute stroke, DWI-trained models consistently outperformed FLAIR-based models, with only modest gains from multimodal combinations. In chronic stroke, increasing training set size improved performance, with diminishing returns beyond several hundred cases. Lesion volume was a key determinant of accuracy: smaller lesions were harder to segment, and models trained on restricted volume ranges generalised poorly. MRI image quality further constrained generalisability: models trained on lower-quality scans transferred poorly, whereas those trained on higher-quality data generalised well to noisier images. Discrepancies between predictions and reference masks were often attributable to limitations in manual annotations. Together, these findings show that automated lesion segmentation can approach human-level performance while identifying key factors governing generalisability and informing the development of lesion segmentation tools.

Many areas in science and engineering now have access to technologies that enable the rapid collection of overwhelming data volumes. While these datasets are vital for understanding phenomena from physical to biological and social systems, the sheer magnitude of the data makes even simple storage, transmission, and basic processing highly challenging. To enable efficient and accurate execution of these data processing tasks, we require new dimensionality reduction tools that 1) do not need expensive, time-consuming training, and 2) preserve the underlying geometry of the data that has the information required to understand the measured system. Specifically, the geometry to be preserved is that induced by the fact that in many applications, streaming high-dimensional data evolves on a low-dimensional attractor manifold. Importantly, we may not know the exact structure of this manifold a priori. To solve these challenges, we present randomized filtering (RF), which leverages a specific instantiation of randomized dimensionality reduction to provably preserve non-linear manifold structure in the embedded space while remaining data-independent and computationally efficient. In this work we build on the rich theoretical promise of randomized dimensionality reduction to develop RF as a real, practical approach. We introduce novel methods, analysis, and experimental verification to illuminate the practicality of RF in diverse scientific applications, including several simulated and real-data examples that showcase the tangible benefits of RF.

Sweepstakes reproduction refers to a highly skewed individual recruitment success without involving natural selection and may apply to individuals in broadcast spawning populations characterised by Type III survivorship. We consider an extension of the model of sweepstakes reproduction for a haploid panmictic population of constant size $N$; the extension also works as an alternative to the Wright-Fisher model. Our model incorporates an upper bound on the random number of potential offspring (juveniles) produced by a given individual. Depending on how the bound behaves relative to the total population size, we obtain the Kingman coalescent, an incomplete Beta-coalescent, or the (complete) Beta-coalescent. We argue that applying such an upper bound is biologically reasonable. Moreover, we estimate the error of the coalescent approximation. The error estimates reveal that convergence can be slow, and small sample size can be sufficient to invalidate convergence, for example if the stated bound is of the form $N/\log N$. We use simulations to investigate the effect of increasing sample size on the site-frequency spectrum. When the limit is a Beta-coalescent, the site frequency spectrum will be as predicted by the limiting tree even though the full coalescent tree may deviate from the limiting one. When in the domain of attraction of the Kingman coalescent the effect of increasing sample size depends on the effective population size as has been noted in the case of the Wright-Fisher model. Conditioning on the population ancestry (the random ancestral relations of the entire population at all times) may have little effect on the site-frequency spectrum for the models considered here (as evidenced by simulation results).

Accumulation of amyloid beta proteins is a defining feature of Alzheimer's disease, and is usually accompanied by cerebrovascular pathology. Evidence suggests that amyloid beta and cerebrovascular pathology are mutually reinforcing; in particular, amyloid beta suppresses perfusion by constricting capillaries, and hypoperfusion promotes the production of amyloid beta. Here, we propose a whole-brain model coupling amyloid beta and blood vessel through a hybrid model consisting of a reaction-diffusion system for the protein dynamics and porous-medium model of blood flow within and between vascular networks: arterial, capillary and venous. We discretize the resulting parabolic--elliptic system of PDEs by means of a high-order discontinuous Galerkin method in space and an implicit Euler scheme in time. Simulations in realistic brain geometries demonstrate the emergence of multistability, implying that a sufficiently large pathogenic protein seeds is necessary to trigger disease outbreak. Motivated by the "two-hit vascular hypothesis" of Alzheimer's disease that hypoperfusive vascular damage triggers amyloid beta pathology, we also demonstrate that localized hypoperfusion, in response to injury, can destabilize the healthy steady state and trigger brain-wide disease outbreak.

The coupled Arctic system is in rapid transition and is set to undergo further dramatic changes over the coming decades. These changes will lead most likely to an ice-free ocean in summer, expected before mid-century. The Arctic will become more strongly influenced by atmospheric and oceanographic processes characteristic of mid-latitudes, increasing the prevalence of contaminants and new biological species. This ongoing transition of the Arctic to a new state necessitates systematic monitoring of all sentinels (variables that make an essential contribution to characterizing the Earth's state) to improve our understanding of the system, enhance forecasting and support knowledge-based decisions. Here, we describe a sustained multi-decadal observation program to be implemented on the Tara Polar Station between 2026 and 2046. The monitoring program is designed as a series of year-long drift expeditions, called Tara Polaris, in the central Arctic Ocean, covering all seasons. The multidisciplinary data will bridge ecological, geochemical, biological, and physical parameters and processes in the atmosphere, sea ice and ocean. In addition, data collected with consistent methodologies over a 20-year period will make it possible to distinguish long-term trends from seasonal and interannual variability. In this paper, we discuss specific measurement challenges in each compartment (i.e., atmosphere, sea ice and ocean) along key sentinels and the most pressing scientific questions to be addressed. The expected outcomes of the Tara Polaris program will enable us to understand and quantify the main feedbacks of the coupled Arctic system, with their seasonal and interannual trends and spatial variability.

Selective attention can momentarily alter visual appearance, but can such effects be learned? We tested whether training attention under sensory competition produces lasting changes in perceived contrast. Across seven days, participants trained on an orientation task with a fixed target location, with or without a salient distractor. Before and after training, we measured the point of subjective equality (PSE). Training under competition produced a reliable push-pull shift. Stimuli at the trained location appeared higher in contrast, whereas stimuli at the untrained location appeared lower. Conversely, training without distractors improved performance but did not alter appearance. Crucially, these opponent shifts were robust to task variations, persisting even in equality judgments designed to minimize response bias. Furthermore, the effect generalized to stimuli with novel orientation and contrast levels. These findings demonstrate that resolving sensory competition does not merely improve discrimination, but durably recalibrates the subjective appearance of the visual world.

Dietary flavonoids associate with disease prevention in epidemiological studies, yet their polypharmacological mechanisms remain unclear. We establish network pharmacology as a systematic framework to characterize flavonoid therapeutic properties through integrated computational, experimental, and epidemiological validation. We constructed a master network of 17,869 human proteins, 14 dietary flavonoids, and 1,496 FDA-approved drugs with 278,768 interactions. Flavonoids averaged 45.3 target proteins per compound compared to 16.8 for FDA-approved drugs (2.7-fold higher; p=7.5x10^-4), reflecting multi-target architecture. Statistical analysis revealed that 71.4% of flavonoids targeted proteins associated with cardiovascular drugs and 78.6% aligned with antineoplastic drug targets. MTT-based Jurkat cell assays confirmed network predictions: high-association flavonoids (luteolin LC50=31.4 microM, myricetin=29.5 microM) produced strong cytotoxicity, while low-association flavonoids showed minimal activity (LC50>200 microM). Network-predicted association strengths correlated with experimental bioactivity (Pearson r=0.918; R^2=0.843). We translated network associations into food-level predictions across 506 foods, identifying 685 food-drug therapeutic combinations. Systematic literature searches confirmed 96 associations supported by 132 unique references. Cardiovascular domains achieved 47.1% validation. Top-validated foods included tea (31 evidence items), blueberries (18 items), tomato (13 items), grape juice (10 items), and plum (9 items). Network pharmacology characterizes dietary polypharmacological properties and generates evidence-based food-therapeutic predictions, bridging nutritional science and systems pharmacology.

String breaking, the process by which flux tubes fragment into hadronic states, is a hallmark of confinement in strongly-interacting quantum field theories. We examine a suite of quantum complexity measures using Matrix Product States to dissect the string breaking process in the 1+1D Schwinger model. We demonstrate the presence of nonlocal quantum correlations along the string that may affect fragmentation dynamics, and show that entanglement and magic offer complementary perspectives on string formation and breaking beyond conventional observables.

Classical low-density parity-check (LDPC) codes are a widely deployed and well-established technology, forming the backbone of modern communication and storage systems. It is well known that, in this classical setting, increasing the girth of the Tanner graph while maintaining regular degree distributions leads simultaneously to good belief-propagation (BP) decoding performance and large minimum distance. In the quantum setting, however, this principle does not directly apply because quantum LDPC codes must satisfy additional orthogonality constraints between their parity-check matrices. When one enforces both orthogonality and regularity in a straightforward manner, the girth is typically reduced and the minimum distance becomes structurally upper bounded. In this work, we overcome this limitation by using permutation matrices with controlled commutativity and by restricting the orthogonality constraints to only the necessary parts of the construction, while preserving regular check-matrix structures. This design breaks the conventional trade-off between orthogonality, regularity, girth, and minimum distance, allowing us to construct quantum LDPC codes with large girth and without the usual distance upper bounds. As a concrete demonstration, we construct a girth-8, (3,12)-regular $[[9216,4612, \leq 48]]$ quantum LDPC code and show that, under BP decoding combined with a low-complexity post-processing algorithm, it achieves a frame error rate as low as $10^{-8}$ on the depolarizing channel with error probability $4 \%$.

Bell measurements (BMs) are ubiquitous in quantum information and technology. They are basic elements for quantum commmunication, computation, and error correction. In particular, when performed on logical qubits encoded in physical photonic qubits, they allow for a read-out of stabilizer syndrome information to enhance loss tolerance in qubit-state transmission and fusion. However, even in an ideal setting without photon loss, BMs cannot be done perfectly based on the simplest experimental toolbox of linear optics. Here we demonstrate that any logical BM on stabilizer codes can always be mapped onto a single physical BM perfomed on any qubit pair from the two codes. As a necessary condition for the success of a logical BM, this provides a general upper bound on its success probability, especially ruling out the possibility that the stabilizer information obtainable from only partially succeeding, physical linear-optics BMs could be combined into the full logical stabilizer information. We formulate sufficient criteria to find schemes for which a single successful BM on the physical level will always allow to obtain the full logical information by suitably adapting the subsequent physical measurements. Our approach based on stabilizer group theory is generally applicable to any stabilizer code, which we demonstrate for quantum parity, five-qubit, standard and rotated planar surface, tree, and seven-qubit Steane codes. Our schemes attain the general upper bound for all these codes, while this bound had previously only been reached for the quantum parity code.

Black hole entropy is identified with the counting of the dynamical degrees of freedom of trapped gravitational modes continually sourced by the Hawking-Unruh process. In the context of linear perturbations of Schwarzschild spacetime the density of states is derived from the orthogonality of states in the solution space of the Regge-Wheeler-Zerilli equation. The otherwise divergent energy and entropy is cutoff by the Planck scale closest approach of constantly accelerating observers near the horizon. The thermal distribution of the trapped modes, which represent shape fluctuations in the near horizon geometry, store a significant fraction of the spacetime mass as observed from far away. Unlike quasi-normal modes the modes are not directly observable outside of $\sim 3 M$ but, being external to the horizon, they affect the propagation of null rays near the black hole. The characteristic frequencies, around 100 Hz for solar mass black holes, are discussed in relation to possible observations.

As is well known, near-horizon (equivalently high acceleration) observers in spherically symmetric black hole spacetimes have a particularly simple form of the quasi-local energy. Using this energy and indistinguishable area quanta satisfying quantum statistics a statistical mechanical description of the Schwarzschild black hole geometry for uniformly accelerating observers is developed. The resulting model has several phases including one with highly excited states, Bose-Einstein condensates, condensates distinct from the usual Bose gas, and degenerate Fermi gases. In the large area limit, relevant for comparison to the Bekenstein-Hawking entropy, the new condensed state is favored over Bose-Einstein condensation and the degenerate Fermi gas. The entropies of the phases, and the entropy of mixing, are computed. The resulting low-entropic condensed state, where the quanta are essentially all in the lowest Bose energy state, provides the framework for the quantization of near-horizon geometric fluctuations, which is explored in a companion paper.

Bosonic codes constitute a promising route to fault-tolerant quantum computing. {Existing Floquet protocols enable analytical construction of bosonic codes but typically rely on slow adiabatic ramps with thousands of driving periods.} In this work, we circumvent this bottleneck by introducing an analytical and deterministic Floquet method that directly synthesizes arbitrary unitaries within a single period. The phase-space unitary ensembles generated by our approach reproduce the Haar-random statistics, enabling practical pseudorandom unitaries in continuous-variable systems. We prepare various prototypical bosonic codes from vacuum and implement single-qubit logical gates with high fidelities using quantum lattice gates. By harnessing the full intrinsic nonlinearity of Josephson junctions, quantum lattice gates decompose quantum circuits into primitive operations for efficient continuous-variable quantum computing.

As quantum devices continue to improve in scale and precision, a central challenge is how to effectively utilize noisy hardware for meaningful computation. Most existing approaches aim to recover noiseless circuit outputs from noisy ones through error mitigation or correction. Here, we show that noisy quantum devices can be directly leveraged as computational resources to enhance the classical simulation of quantum circuits. We introduce the Noisy-device-enhanced Classical Simulation (NDE-CS) protocol, which improves stabilizer-based classical Monte Carlo simulation methods by incorporating data obtained from noisy quantum hardware. Specifically, NDE-CS uses noisy executions of a target circuit together with noisy Clifford circuits to learn how the target circuit can be expressed in terms of Clifford circuits under realistic noise. The same learned relation can then be reused in the noiseless Clifford limit, enabling accurate estimation of ideal expectation values with substantially reduced sampling cost. Numerical simulations on Trotterized Ising circuits demonstrate that NDE-CS achieves orders-of-magnitude reductions in sampling cost compared to the underlying purely classical Monte Carlo approaches from which it is derived, while maintaining the same accuracy. We also compare NDE-CS with Sparse Pauli Dynamics (SPD), a powerful classical framework capable of simulating quantum circuits at previously inaccessible scales, and provide an example where the cost of SPD scales exponentially with system size, while NDE-CS scales much more favorably. These results establish NDE-CS as a scalable hybrid simulation approach for quantum circuits, where noise can be harnessed as a computational asset.

We obtain the metric which describes the spacetime corresponding to the Frolov black hole in the presence of a cloud of strings and discuss how this cloud affects the regularity of the solution and the energy conditions. In addition, we analyze geodesics, effective potential, and several thermodynamic aspects. Finally, we compare our results with the corresponding findings in the literature for the original Frolov black hole, that is, in the absence of a cloud of strings

We prove that $\mathrm{deg}(f) \leq 2 \, \mathrm{rdeg}(f)^4$ for every Boolean function $f$, where $\mathrm{deg}(f)$ is the degree of $f$ and $\mathrm{rdeg}(f)$ is the rational degree of $f$. This resolves the second of the three open problems stated by Nisan and Szegedy, and attributed to Fortnow, in 1994.

We study Krylov complexity for quantum systems whose Hamiltonians factorise as tensor products. We prove that complexity is superadditive under tensor products, $C_{12}\ge C_1+C_2$, and identify a positive operator that quantifies the resulting excess complexity. The underlying mechanism is made transparent by introducing a Krylov graph representation in which tensor products generate a higher-dimensional lattice whose diagonal shells encode operator growth and binomial path multiplicities. In the continuum limit, Krylov dynamics reduces to diffusion on this graph, with superadditivity arising from geometric broadening across shells. Explicit examples illustrate how deviations from synchronous evolution generate bounded, oscillatory excess complexity.

For a given quantum state used in sensing, the quantum Cramér-Rao bound (QCRB) sets a fundamental limit on the precision achievable by an unbiased estimator of an unknown parameter, determined by the inverse of the quantum Fisher information (QFI). The QFI serves as an upper bound on the classical Fisher information (CFI), representing the maximum extractable information about the unknown parameter from measurements on a physical system. Thus, a central goal in quantum parameter estimation is to find a measurement, described by a POVM, that saturates the QFI (achieves maximum CFI), and thereby achieves the QCRB. In the idealization that one uses pure states and unitary encodings for sensing, discontinuities can appear in the CFI but not the QFI. In this article, we demonstrate that these discontinuities are important features, quantifying how much Fisher information is lost in the presence of noise. We refer to this as the Fisher information "fragility". We present a simple framework for understanding how discontinuities increase fragility through Jensen's inequality, and demonstrate how one can use this framework to design more robust POVMs for quantum advantage in metrology.

Computing functional determinants of differential operators is central to any field-theoretical calculation relying on a saddle-point expansion. A variety of approaches is available for the computation that avoid having to know the eigenspectrum of the operator, and in particular the Gel'fand-Yaglom theorem and the Green's function method. In this note, we show how both approaches can be constructed using a contour integral argument and conclude that these are completely equivalent for computing ratios of determinants of one-dimensional operators. Furthermore, we comment on the presence of vanishing as well as negative eigenvalues and show how the Green's function method provides a natural prescription for handling them.

We compute the full one-loop corrections to the primordial tensor power spectrum in an inflationary scenario with a non-minimally coupled spectator field, using the in-in formalism. We derive semi-analytic results for the scalar-sourced one-loop tensor spectrum and the effective tensor-to-scalar ratio, $r_{\mathrm{eff}}$ . We consider two representative coupling functions: a localized Gaussian dip (Model G), which leads to moderate loop corrections, and a rapidly oscillatory coupling (Model O), which can yield much larger loop contributions. For Model G, we find a $\mathcal{O}(1)$ correction to $r_{\mathrm{eff}}$ while Model O can significantly enhance $r_{\mathrm{eff}}$ by several orders of magnitude (relative to the tree-level value). We further calculate the energy density of primordial gravitational waves. Assuming that primordial black holes with mass $10^{-12}M_{\odot}$ generated in this scenario, constitute all of the dark matter, we find that the results are several orders of magnitude lower than the sensitivities of Taiji/TianQin/LISA.

Modern Quantum optics largely remains quasi-stationary, far from intrinsic optical field timescales. Ultrafast quantum optics seeks to generate, shape, and measure quantum states of light on femtosecond and attosecond timescales. Here we introduce a quantum light field squeezer (QLFS) that enables the generation and attosecond control of ultrafast broadband squeezed light. Using degenerate four-wave mixing in a quasi-collinear focusing geometry, our approach overcomes conventional broadband phase-matching limits, producing intensity- and phase-squeezed states directly from few-cycle laser pulses. Our ultrafast quantum optical metrology reveals a time-dependent squeezing distribution across individual half-cycles of the electric field. Incorporating this time-dependent squeezing into strong-field simulations shows that the temporal redistribution of quantum uncertainty reshapes the high-harmonic emission. Moreover, by tuning the relative pulse delay and phase-matching angle, we achieve attosecond precision in controlling the squeezing characteristics by visualizing inferred effective Wigner representations of the quantum light field. Beyond characterization, we demonstrate that the quantum light-induced tunneling-current noise is sensitive to the nonclassical intensity-noise statistics of the driving squeezed light, with sub-femtosecond control. Together, these results extend the generation, control, and effective phase-space representation of squeezed light into the ultrafast and attosecond regime, opening new avenues for quantum optics in strong-field and solid-state systems.

Quantum low-density parity-check (QLDPC) codes provide a practical balance between error-correction capability and implementation complexity in quantum error correction (QEC). In this paper, we propose an algebraic construction based on dyadic matrices for designing both classical and quantum LDPC codes. The method first generates classical binary quasi-dyadic LDPC codes whose Tanner graphs have girth 6. It is then extended to the Calderbank-Shor-Steane (CSS) framework, where the two component parity-check matrices are built to satisfy the compatibility condition required by the recently introduced CAMEL-ensemble quaternary belief propagation decoder. This compatibility condition ensures that all unavoidable cycles of length 4 are assembled in a single variable node, allowing the mitigation of their detrimental effects by decimating that variable node.

We construct a novel exact solution of the Einstein-scalar-Maxwell equations describing a dynamical black hole immersed in an external, time-dependent electromagnetic field. Motivated by the need for more realistic analytical black hole models, our construction incorporates two key ingredients often neglected in exact solutions: a fully dynamical cosmological background and the non-perturbative backreaction of external electromagnetic fields. The compact object is obtained by dressing a Schwarzschild black hole with a radially and temporally dependent scalar field, yielding a time-dependent generalization of the Fisher-Janis-Newman-Winicour solution within the Fonarev framework. The external electromagnetic field is generated via a Lie point symmetry of the Einstein-scalar-Maxwell system, which exports the effect of a Harrison transformation to dynamical settings provided a spacelike Killing vector is present. The resulting spacetime combines a spherically symmetric dynamical horizon with an axisymmetric electromagnetic field and exhibits a rich asymptotic structure mixing Friedmann-Lemaître-Robertson-Walker and Levi-Civita geometries. We show that the time dependence of the configuration plays a crucial role in potentially cloaking curvature singularities, which would otherwise be generically naked in the stationary limit. We analyze the geometric and physical properties of the solution, including its asymptotic behavior, algebraic classification, and the structure of trapped surfaces defining the dynamical horizon. Possible implications for primordial black holes and some astrophysical applications, as well as extensions to higher dimensions, are also discussed.

We construct a $G$-cluster Hamiltonian in 2+1 dimensions and analyze its properties. This model exhibits a $G\times2\mathrm{Rep}(G)$ symmetry, where the $2\mathrm{Rep}(G)$ sector realizes a non-invertible symmetry obtained by condensing appropriate algebra objects in $\mathrm{Rep}(G)$. Using the symmetry interpolation method, we construct $S^1$- and $S^2$-parameterized families of short-range-entangled (SRE) states by interpolating an either invertible $0$-form or $1$-form symmetry contained in $G\times2\mathrm{Rep}(G)$. Applying an adiabatic evolution argument to this family, we analyze the pumped interface mode generated by this adiabatic process. We then explicitly construct the symmetry operator acting on the interface and show that the interface mode carries a nontrivial charge under this symmetry, thereby demonstrating the nontriviality of the parameterized family.

We construct 2+1-dimensional lattice models of symmetry-protected topological (SPT) phases with non-invertible symmetries and investigate their properties using tensor networks. These models, which we refer to as generalized cluster models, are constructed by gauging a subgroup symmetry $H \subset G$ in models with a finite group 0-form symmetry $G$. By construction, these models have a non-invertible symmetry described by the group-theoretical fusion 2-category $\mathcal{C}(G; H)$. After identifying the tensor network representations of the symmetry operators, we study the symmetry acting on the interface between two generalized cluster states. In particular, we will see that the symmetry at the interface is described by a multifusion category known as the strip 2-algebra. By studying possible interface modes allowed by this symmetry, we show that the interface between generalized cluster states in different SPT phases must be degenerate. This result generalizes the ordinary bulk-boundary correspondence. Furthermore, we construct parameterized families of generalized cluster states and study the topological charge pumping phenomena, known as the generalized Thouless pump. We exemplify our construction with several concrete cases, and compare them with known phases, such as SPT phases with $2\mathrm{Rep}((\mathbb{Z}_{2}^{[1]}\times\mathbb{Z}_{2}^{[1]})\rtimes\mathbb{Z}_{2}^{[0]})$ symmetry.

Entanglement swapping is a key primitive for distributing entanglement across nodes in quantum networks. In standard protocols, the outcome of the intermediate measurement determines the resulting state, making the process inherently probabilistic and requiring postselection. In this work, we fully characterize those measurements under which entanglement swapping becomes deterministic: for arbitrary pure inputs, every measurement outcome produces local-unitarily equivalent states. We also show that an optimal measurement, maximizing a concurrence-type entanglement measure, is built from complex Hadamard matrices. For this optimal protocol, we provide a complete, dimension-dependent classification of deterministic entanglement-swapping measurements: unique in dimensions $d=2,3$, infinite for $d=4$, and comprising $72$ inequivalent classes for $d=5$. We further consider a general network with multiple swapping nodes and show that, for $d=2,3$ the resulting end-to-end state is independent of the order in which the repeaters perform the optimal measurements. Our results establish optimal entanglement-swapping schemes that are post-selection free, in the sense that they distribute entanglement across generic quantum network architectures without unfavorable measurement outcomes.

Although the principle of superposition lies at the heart of quantum mechanics and is the root of almost all quantum phenomena such as coherence and entanglement, its quantification, except for that related to the resource theory of coherence and interference, remains relatively less studied. In this work, we address quantification of superposition from an information-theoretic perspective. We introduce a family of quantifiers of superposition, the phase-sensitive superposition, by taking into account the phases of amplitudes in the superposition of a fixed basis states (e.g., computational basis states). We establish a conservation relation for the phase-sensitive superposition, which is a kind of complementary relation and is reminiscent of wave-particle duality. We evaluate explicitly the second moment of phase-sensitive superposition and show that it is intrinsically related to the $l^2$-norm coherence. We characterize the dephasing channel induced by the maximally superposed states. We investigate the minimum and maximum superpositions, reveal their basic properties, and illustrate them through various examples. We further explore the dynamics of superposition in the Grover search algorithm, and demonstrate a complementary relation between superposition and success probability of the search algorithm. These results and quantifiers offer tools for analyzing structural features and implications of quantum superposition.

This whitepaper surveys the current landscape and short- to mid-term prospects for quantum-enabled optimization and machine learning use cases in industrial settings. Grounded in the QCHALLenge program, it synthesizes hardware trajectories from different quantum architectures and providers, and assesses their maturity and potential for real-world use cases under a standardized traffic-light evaluation framework. We provide a concise summary of relevant hardware roadmaps, distinguishing superconducting and ion-trap technologies, their current states, modalities, and projected scaling trajectories. The core of the presented work are the use case evaluations in the domains of optimization problems and machine learning applications. For the conducted experiments, we apply a consistent set of evaluation criteria (model formulation, scalability, solution quality, runtime, and transferability) which are assessed in a shared system of three categories, ranging from optimistic (solutions produced by quantum computers are competitive with classical methods and/or a clear path to a quantum advantage is shown) to pessimistic (significant hurdles prevent practical application of quantum solutions now and potentially in the future). The resulting verdicts illuminate where quantum approaches currently offer promise, where hybrid classical-quantum strategies are most viable, and where classical methods are expected to remain superior.

This paper introduces a framework for constructing Calderbank-Shor-Steane (CSS) codes that support fault-tolerant logical transversal $Z$-rotations. Using this framework, we obtain asymptotically good CSS codes that fault-tolerantly realize the logical transversal Clifford group. Furthermore, investigating CSS-T codes, we: (a) demonstrate asymptotically good CSS-T codes wherein the transversal $T$ realizes the logical transversal $S^{\dagger}$; (b) show that the condition $C_2 \ast C_1 \subseteq C_1^{\perp}$ is necessary but not sufficient for CSS-T codes; and (c) revise the characterizations of CSS-T codes wherein the transversal $T$ implements the logical identity and the logical transversal $T$, respectively.

We present systematic and resource-efficient constructions of continuous symmetry projectors, particularly $U(1)$ particle number and $SU(2)$ total spin, tailored for fault-tolerant quantum computations. Our approach employs a linear combination of unitaries (LCU) as well as generalized quantum signal processing (GQSP and GQSVT) to implement projectors. These projectors can then be coherently applied as state filters prior to quantum phase estimation (QPE). We analyze their asymptotic gate complexities for explicit circuit realizations. For the particle number and $S_z$ symmetries, GQSP offers favorable resource usage features owing to its low ancilla qubit requirements and robustness to finite precision rotation gate synthesis. For the total spin projection, the structured decomposition of $\hat{P}_{S,M_S}$ reduces the projector T gate count. Numerical simulations show that symmetry filtering substantially increases the QPE success probability, leading to a lower overall cost compared to that of unfiltered approaches across representative molecular systems. Resource estimates further indicate that the cost of symmetry filtering is $3$ to $4$ orders of magnitude lower than that of the subsequent phase estimation step This advantage is especially relevant in large, strongly correlated systems, such as FeMoco, a standard strongly correlated open-shell benchmark. For FeMoco, the QPE cost is estimated at ${\sim}10^{10}$ T gates, while our symmetry projector requires only ${\sim}10^{6}$--$10^{7}$ T gates. These results establish continuous-symmetry projectors as practical and scalable tools for state preparation in quantum chemistry and provide a pathway toward realizing more efficient fault-tolerant quantum simulations.

We present a new empirical relation between the standardized magnitude ($m$) of Type Ia supernovae (SNe Ia) and redshift ($z$). Using Pantheon+ and DES-SN5YR, we find a negative linear correlation between $m-5\log(z(1+z))$ and $z$, implying that their magnitude--redshift relation can be parametrized with just two parameters: an intercept $\mathcal{M}$ and a slope $b$. This relation corresponds to the luminosity distance $d_L(z)=c\,H_0^{-1}z(1+z)10^{bz/5}$ and is valid up to at least $z\simeq1.1$. It outperforms the $Λ$CDM and flat $w$CDM models and the (2,1) Padé approximant for $d_L(z)$, and performs comparably to the flat $Λ$CDM model and the (2,1) Padé($j_0=1$) model of Hu et al. Furthermore, the relation is stable in the absence of low-$z$ SNe, making it suitable for fitting Hubble diagrams of SNe Ia without the need to add a low-$z$ sample. In deep fields in particular, assuming that the large-scale density is independent of the comoving radial coordinate, $b\propto q_0+1$. We fit the empirical relation to SN data in eight deep-field regions and find that their fitted $\mathcal{M}$ and $b$ parameters are consistent within $1.6\,σ$, in agreement with isotropy. The inferred $q_0$ values, ranging from $-0.6$ to $-0.4$, are consistent within $1.5\,σ$ and significantly lower than zero, indicating statistically consistent cosmic acceleration across all eight regions. We apply the empirical relation to the DES-Dovekie and Amalgame SN samples, finding $b$ values consistent with those from DES-SN5YR and Pantheon+. Finally, using the empirical relation in the hemispheric comparison method applied to Pantheon+ up to $z=1.1$, we find no evidence for anisotropies in $\mathcal{M}$ and $b$.

Neutral atom Quantum Processing Units (QPUs) are emerging as a popular quantum computing technology due to their large qubit counts and flexible connectivity. However, performance challenges arise as large circuits experience significant fidelity drops, while small circuits underutilize hardware and face initialization latency issues. To tackle these problems, we propose $\textit{multi-programming on neutral atom QPUs}$, allowing the co-execution of multiple circuits by logically partitioning the qubit array. This approach increases resource utilization and mitigates initialization latency while maintaining result fidelity. Currently, state-of-the-art compilers for neutral atom architectures do not support multi-programming. To fill this gap, we introduce MultiQ, the first system designed for this purpose. MultiQ addresses three main challenges: (i) it compiles circuits into a $\textit{virtual zone layout}$ to optimize spatio-temporal hardware utilization; (ii) it parallelizes the execution of co-located circuits, allowing single hardware instructions to operate on different circuits; and (iii) it includes an algorithm to verify the functional independence of the bundled circuits. MultiQ functions as a cross-layer system comprising a compiler, controller, and checker. Our compiler generates \emph{virtual zone layouts} to enhance performance, while the controller efficiently maps these layouts onto the hardware and resolves any conflicts. The checker ensures the correct bundling of circuits. Experimental results show a throughput increase from 3.8$\times$ to 12.3$\times$ when multi-programming 4 to 14 circuits, with fidelity largely maintained, ranging from a 1.3% improvement for four circuits to only a 3.5% loss for fourteen circuits. Overall, MultiQ facilitates concurrent execution of multiple quantum circuits, boosting throughput and hardware utilization.

Geometry is wavy: even at the purely geometric level (no particular theory chosen), curvature satisfies a covariant quasilinear wave equation. In Riemannian geometry equipped with the Levi-Civita connection, the Riemann curvature tensor obeys a wave equation of the schematic form \[ \Box Riem=\mathcal{Q}(Riem,Riem), \] where $\mathcal{Q}(Riem,Riem)$ denotes the terms quadratic in the curvature arising from the Bianchi identities. In this work, we generalize this curvature wave equation to spacetimes endowed with a generic affine connection possessing torsion and nonmetricity. Working within the metric-affine framework, we derive the corresponding wave equation for the Riemann tensor and analyze its structure in several geometrically and physically distinguished settings, including Einstein spaces, teleparallel gravity, and Einstein-Cartan theory.

Scalable trapped-ion quantum computing requires fast and reliable transport of ions through complex, segmented radiofrequency trap architectures without inducing excessive motional excitation. We present a numerical toolchain for the systematic generation of time-dependent electrode voltages enabling fast, low-excitation ion shuttling in segmented radiofrequency traps. Based on a model of the trap electrode geometry, the framework combines an electrostatic field solver, efficient unconstrained optimization, waveform postprocessing, and dynamical simulations of ion motion to compute voltage waveforms that realize prescribed transport trajectories while respecting experimental constraints such as voltage limits and bandwidth. The toolchain supports arbitrary trap geometries, including junctions and multi-zone layouts, and allows for the flexible incorporation of optimization objectives. We provide a detailed assessment of the accuracy of the framework by investigating its numerical stability and by comparing measured and predicted secular frequencies. The framework is optimized for numerical performance, enabling rapid numerical prototyping of trap architectures of increasing complexity. As application examples, we apply the framework to the transport of a potential well along a linear, uniformly segmented trap, and we compute a solution for shuttling a potential well around the corner of an X-type trap junction. The presented approach provides an extensible and highly efficient numerical foundation for designing and validating transport protocols in current and next-generation trapped-ion processors.

We study a six-dimensional Kaluza-Klein theory with spacetime topology $M_4 \times S^2$ and analyze the gauge sector arising from dimensional reduction. Using normalized Killing vectors on $S^2$, we explicitly construct the reduced Yang-Mills action and determine the corresponding gauge kinetic matrix. Despite the $SO(3)$ isometry of $S^2$, we show that only two physical gauge fields propagate in four dimensions. The gauge kinetic matrix therefore has rank two and possesses a single zero eigenvalue. We demonstrate that this degeneracy is a direct consequence of the coset structure $S^2 \simeq SO(3)/SO(2)$ and reflects a non-dynamical gauge direction rather than an inconsistency of the reduction. Our results clarify the geometric origin of gauge degrees of freedom in Kaluza-Klein reductions on coset spaces.

Practical QKD modulators introduce correlations between consecutively emitted pulses due to bandwidth limitations, violating key assumptions underlying many security proof techniques. Here, we address this problem by introducing a simple yet powerful mathematical framework to directly extend phase-error-estimation-based security proofs for imperfect but uncorrelated sources to also incorporate encoding correlations. Our framework overcomes important limitations of previous approaches in terms of generality and rigor, significantly narrowing the gap between theoretical security guarantees and real-world QKD implementations.

We present a semi-device-independent quantum random number generator (QRNG) based on the violation of a contextuality inequality, implemented by the integration of two silicon photonic chips. Our system combines a heralded single-photon source with a reconfigurable interferometric mesh to implement qutrit state preparation, transformations, and measurements suitable for testing a KCBS contextuality inequality. This architecture enables the generation of random numbers from the intrinsic randomness of single-photon interference in a complex optical network, while simultaneously allowing a quantitative certification of their security without requiring entanglement. We observe a contextuality violation exceeding the classical bound by more than 10σ, unambiguously confirming non-classical behavior. From this violation, we certify a conditional min-entropy per experimental round of Hmin = 0.077 +- 0.002, derived via a tailored semidefinite-programming-based security analysis. Each measurement outcome therefore contains at least 0.077 +- 0.002 bits of extractable genuine randomness, corresponding to an asymptotic generation rate of 21.7 +- 0.5 bits/s. These results establish a viable route towards general-purpose, untrusted quantum random number generators compatible with practical integrated photonic quantum networks.

Photonic systems offer a promising platform for interconnecting quantum processors and enabling scalable, networked architectures. Designing and verifying such architectures requires a unified formalism that integrates linear algebraic reasoning with probabilistic and control-flow structures. In this work, we introduce a graphical framework for distributed quantum computing that brings together linear optics, the ZX-calculus, and dataflow programming. Our language supports the formal analysis and optimization of distributed protocols involving both qubits and photonic modes, with explicit interfaces for classical control and feedforward, all expressed within a synchronous dataflow model with discrete-time dynamics. Within this setting, we classify entangling photonic fusion measurements, show how their induced Pauli errors can be corrected via a novel flow structure for fusion networks, and establish correctness proofs for new repeat-until-success protocols enabling arbitrary fusions. Layer by layer, we construct qubit architectures incorporating practical optical components such as beam splitters, switches, and photon sources, with graphical proofs that they are deterministic and support universal quantum computation. Together, these results establish a foundation for verifiable compilation and automated optimization in networked quantum computing.

We verify transverse spatial entanglement of photon-pairs generated in spontaneous parametric down conversion using a nonlinear interferometric technique without relying on any coincidence detection. We experimentally demonstrate the violation of the Einstein-Podolsky-Rosen criterion and of the Mancini-Giovannetti-Vitali-Tombesi criterion using single photon interference of one of the photons of the pairs. We also provide a comprehensive theoretical analysis. The experimental results that we have obtained show good agreement with the theoretical values. Our method performs well under experimental losses and can be applied to highly non-degenerate sources, where there are no suitable detectors for one of the photons in the quantum state and our method could also be extended to the discrete degrees of freedom to certify high-dimensional (OAM) entanglement.

The limit between the two regimes of spontaneous-parametric down-conversion (SPDC) or four-wave-mixing (FWM) regarding the pump intensity has been theoretically investigated using a semi-classical model and analytical calculations. A unitless quantity has been defined, corresponding to the photon-pairs flux per frequency unit: it has been found equal to 0.369 at this limit. The ratio between the magnitudes of the electric fields of the generated photons and of vacuum has been also calculated, equal to 1.718, and the pump intensity has been plotted as a function of the interaction length for different values of the second-order electric susceptibility in the case of SPDC and of the third-order electric susceptibility for FWM. These quantitative results confirm that below the limit, the nonlinear process can be truly considered as spontaneous, i.e. mainly seeded by the quantum fluctuations of vacuum, while the generated photons mainly govern the pump photon splitting above the limit, which corresponds more to an optical parametric amplification / difference frequency generation regime. Knowing quantitatively the limit between the two regimes thanks to the present calculations will be a useful guide for further quantum calculations and measurements from either side of the limit in order to catch the full quantum picture of SPDC and FWM from low to high pump intensities.

Dynamical decoupling (DD) and bath engineering are two parallel techniques employed to mitigate qubit decoherence resulting from their unavoidable coupling to the environment. Here, we present a hybrid DD approach that integrates pulsed DD with bath spin polarization to enhance qubit coherence within the central spin model. This model, which can be realized using GaAs semiconductor quantum dots or analogous quantum simulators, demonstrates a significant extension of the central spin's coherence time by approximately 2 to 3 orders of magnitude that compared with the free-induced decay time, where the dominant contribution from DD and a moderate improvement from spin-bath polarization. This study, which integrates uniaxial dynamical decoupling and auxiliary bath-spin engineering, paves the way for prolonging coherence times in various practical quantum systems, including GaAs/AlGaAs, silicon and Si/SiGe. And this advancement holds substantial promise for applications in quantum information processing.

The exploration of photonic systems for quantum information processing has generated widespread interest in multiple cutting-edge research fields. Photonic frequency encoding stands out as an especially viable approach, given its natural alignment with established optical communication technologies, including fiber networks and wavelength-division multiplexing systems. Substantial reductions in hardware resources and improvements in quantum performance can be expected by utilizing multiple frequency modes. The integration of nonlinear photonics with microresonators provides a compelling way for generating frequency-correlated photon pairs across discrete spectral modes. Here, by leveraging the high material nonlinearity and low nonlinear loss, we demonstrate an efficient chip-scale multi-wavelength quantum light source based on AlGaAs-on-insulator, featuring a free spectral range of approximately 200 GHz at telecom wavelengths. The optimized submicron waveguide geometry provides both high effective nonlinearity (~550 m$^{-1}$W$^{-1}$) and broad generation bandwidth, producing eleven distinct wavelength pairs across a 35.2 nm bandwidth with an average spectral brightness of 2.64 GHz mW$^{-2}$nm$^{-1}$. The generation of energy-time entanglement for each pair of frequency modes is verified through Franson interferometry, yielding an average net visibility of 93.1%. With its exceptional optical gain and lasing capabilities, the AlGaAs-on-insulator platform developed here shows outstanding potential for realizing fully integrated, ready-to-deploy quantum photonic systems on chip.

The formation channels of binary black hole (BBH) mergers detected by the LIGO/Virgo/KAGRA (LVK) network remain uncertain. While BBH mergers occurring inside active galactic nucleus (AGN) disks may interact with surrounding gas and generate observable optical flares. We test this scenario by quantifying the spatial and temporal correlation between BBH events in GWTC-4.0 and AGN flares identified from six years of the Zwicky Transient Facility (ZTF) DR23 data. Using 80 BBH mergers selected for adequate localization, redshift reach, observing-epoch overlap, and ZTF sky coverage, we construct a likelihood for the flare-associated fraction, $f_{\rm flare}$, that combines each event's 3D localization with a locally estimated flare number density derived from a 3D Voronoi tessellation, while explicitly accounting for survey boundaries and incomplete catalog coverage. Adopting a 200-day post-merger time window for potential counterparts, we infer $f_{\rm flare} = 0.07_{-0.05}^{+0.25}$ (90\% confidence level). This non-zero maximum-likelihood value is driven primarily by GW190412, for which a single flare candidate (J143041.67+355703.8) is consistent in both time and spatial position. Excluding GW190412 yields results consistent with no association and an upper limit of $f_{\rm flare} < 0.17$ at 90\% confidence level. The intrinsic properties of GW190412 and the characteristics of the candidate host AGN are broadly consistent with theoretical expectations for the AGN-disk formation channel, motivating continued, targeted electromagnetic follow-up of well-localized and highly asymmetric BBH mergers in current and upcoming time-domain surveys.

We develop a formalism to compute non-perturbative 5-point scattering amplitudes and apply it to gravitational waveforms in the two-body problem for arbitrary trajectories. Drawing inspiration from Feshbach's projector formalism in nuclear physics, we introduce effective potentials governing graviton emission and relate them to perturbative scattering amplitudes at arbitrary order in the gravitational coupling and mass ratio. Once these potentials are determined, the corresponding non-perturbative amplitudes in the classical limit are obtained by iterative insertions and subsequently translated into gravitational waveforms using the KMOC formalism. As an application, we compute the gravitational waveform emitted by a two-body system moving along a generic, potentially highly bent, trajectory. Importantly, our formalism extends effective field theory matching of the gravitational two-body potential to radiative phenomena, enabling the extraction of gravitational-wave source terms directly from perturbative on-shell scattering amplitudes.

Current quantum secure direct communication (QSDC) protocols guarantee communication security by estimating the error rates of photons in the X and Z bases. This take the reference frame calibration between communicating parties as a necessary prerequisite. However, in mobile communications scenarios, achieving continuous and accurate reference frame calibration poses significant challenges. To address this issue, this paper proposes a reference-frame-independent (RFI) QSDC protocol. This protocol only requires ensuring the calibration accuracy of one direction of the reference frame, while allowing a misalignment angle $β$ in the other two directions. To improve the protocol's robustness against reference frame fluctuations, we introduce a $β$-independent parameter C into the security analysis framework and rederive the protocol's security bounds. Additionally, we construct a system model and optimize the pulse intensity of the signal states, enabling the protocol to achieve optimal performance under each level of channel attenuation. At an attenuation of 10 dB (corresponding to a communication distance of 25 km), the secrecy message capacities for $β= 0^{ \circ} $ and $45^{ \circ} $ are $8.765 \times10^{-6}$ bit/pulse and $4.150 \times10^{-6}$ bit/pulse, respectively. Compared with the single-photon-based QSDC, the communication distance of the protocol proposed in this paper is significantly extended. When $β= 0^{ \circ} $ and $45^{ \circ} $, the maximum transmission distances of the RFI QSDC protocol are 27.875 km and 26.750 km, which is about 155.9 % and 149.7 % of that of the single-photon-based QSDC protocol.

We investigate the quantum structure of black hole interiors in Horava-Lifshitz gravity by analyzing the Wheeler-DeWitt equation in minisuperspace. Focusing on the ultraviolet regime, where higher-order spatial curvature terms dominate, we derive analytical solutions in this UV limit for both the original Horava-Lifshitz action and its analytically continued counterpart. We study their behavior near the event horizon and the classical singularity, with particular attention to the interpretation of the wave function in terms of the annihilation-to-nothing scenario proposed in general relativity. In this paper, we have considered cases in which the two-dimensional spatial section is spherical, planar, or hyperbolic, as well as models with positive, negative, or vanishing cosmological constant. In all cases, we find that the terms dominating in the ultraviolet regime, together with the effects of the running scaling parameter, act to suppress the annihilation-to-nothing behavior. These results suggest that, at least within the range explored in this study, the characteristic annihilation-to-nothing behavior does not appear in the ultraviolet regime of Horava-Lifshitz gravity, and provide a new perspective on the understanding of singularity resolution in quantum gravity.

Identifying the state of a quantum bit (qubit), known as quantum state discrimination, is a crucial operation in quantum computing. However, it has been the most error-prone and time-consuming operation on superconducting quantum processors. Due to stringent timing constraints and algorithmic complexity, most qubit state discrimination methods are executed offline. In this work, we present an enhanced real-time quantum state discrimination system leveraging FPGA-based AI Engine technology. A multi-layer neural network has been developed and implemented on the AMD Xilinx VCK190 FPGA platform, enabling accurate in-situ state discrimination and supporting mid-circuit measurement experiments for multiple qubits. Our approach leverages recent advancements in architecture research and design, utilizing specialized AI/ML accelerators to optimize quantum experiments and reduce the use of FPGA resources.

In this article, we investigate the possibility that the large-scale magnetic fields observed in galaxies, of the order of microgauss, arise naturally from a complex Scalar Field Dark Matter (SFDM) halo charged under a local $U(1)$ symmetry. Extending our previous work, where multistate SFDM solutions were shown to form ``gravitational atoms'' capable of explaining the anisotropic distribution of satellite galaxies (VPOS), we analyze here the coupled dynamics of the scalar and a gauge field at the perturbative level. By solving the perturbed Klein-Gordon and gauge-field equations, we find the temporal evolution and show that the spatial structure of the induced electromagnetic fields is governed by the same spherical Bessel functions and spherical harmonics that characterize the ground and excited states of the multi-state SFDM halo. Remarkably, the presence of the gauge field does not modify the dark-matter density distribution, which preserves the multi-state configuration previously obtained. Our results demonstrate that a charged multi-state SFDM halo can generate coherent, large-scale magnetic fields whose morphology is determined by the excited modes of the scalar field, providing a unified framework in which both galactic magnetic fields and VPOS-like structures originate from the underlying quantum nature of dark matter.

Emerging theoretical frameworks suggest that physical properties of matter can be altered within an optical cavity by harnessing quantum vacuum electromagnetic fluctuations, even in the total absence of external driving fields. Among the most intriguing predictions is the potential to noninvasively manipulate superconductivity. Here, we experimentally observe modified superconductivity in niobium nitride (NbN) thin films within high-quality-factor ($Q$) terahertz cavities. Using terahertz time-domain spectroscopy, we characterize the NbN response both in free space and within a high-$Q$ photonic-crystal cavity. Our analysis reveals significant cavity-induced modifications to the optical conductivity. A theoretical model indicates that these changes originate from a substantial ($\sim13\,\%$) reduction in the superfluid density and a minor ($\sim2\,\%$) reduction in the superconducting gap, driven by cavity vacuum fluctuations. These results demonstrate a platform for engineering ground states via vacuum--matter coupling, opening frontiers in cavity materials science.

Tensor network states are an indispensable tool for the simulation of strongly correlated quantum many-body systems. In recent years, tree tensor network states (TTNS) have been successfully used for two-dimensional systems and to benchmark quantum simulation approaches for condensed matter, nuclear, and particle physics. In comparison to the more traditional approach based on matrix product states (MPS), the graph distance of physical degrees of freedom can be drastically reduced in TTNS. Surprisingly, it turns out that, for large systems in $D>1$ spatial dimensions, MPS simulations of low-energy states are nevertheless more efficient than TTNS simulations. With a focus on $D=2$ and 3, the scaling of computational costs for different boundary conditions is determined under the assumption that the system obeys an entanglement (log-)area law, implying that bond dimensions scale exponentially in the surface area of the associated subsystems.

It is commonly thought that observers in distinct branches of an Everettian multiverse cannot communicate without violating the linearity of quantum theory. Here we show a counterexample, demonstrating that inter-branch communication is in fact possible, entirely within standard quantum theory. We do this by considering a Wigner's-friend scenario, where an observer (Wigner) can have quantum control over another observer (the friend). We present a thought experiment where the friend in superposition can receive a message written by a distinct copy of themselves in the multiverse, with the aid of Wigner. To maintain the unitarity of quantum theory, the observers must have no memory of the message that they sent. Our thought experiment challenges conventional wisdom regarding the ultimate limits of what is possible in an Everettian multiverse. It has a surprising potential application which involves using knowledge-creation paradoxes for testing Everettian quantum theory against single-world theories.

Feedback-based quantum algorithms (FQAs) operate by iteratively growing a quantum circuit to optimize a given task. At each step, feedback from qubit measurements is used to inform the next quantum circuit update. In practice, the sampling cost associated with these measurements can be significant. Here, we ask whether FQA parameter sequences can be predicted using classical machine learning, obviating the need for qubit measurements altogether. To this end, we train a teacher-student model to map a MaxCut problem instance to an associated FQA parameter curve in a single classical inference step. Numerical experiments show that this model can accurately predict FQA parameter curves across a range of problem sizes, including problem sizes not seen during model training. To evaluate performance, we compare the predicted parameter curves in simulation against FQA reference curves and linear quantum annealing schedules. We observe similar results to the former and performance improvements over the latter. These results suggest that machine learning can offer a heuristic, practical path to reducing sampling costs and resource overheads in quantum algorithms.

In this work, we systematically assess the performance of a new method from [H. Shah et al., Phys. Rev. C 113, L012201] for locating the QCD critical point using constant-entropy contours by testing it against various effective QCD approaches. We demonstrate that, while the method yields spurious critical points in purely hadronic models (HRG) due to non-parabolic contour behavior at low temperatures ($T \lesssim 120$ MeV), it accurately reproduces the CP location in frameworks that feature a genuine phase transition and benchmarked against lattice QCD, such as Holographic Einstein-Maxwell-Dilaton, and Functional QCD approaches. Building on our previous determination of constant entropy contours using lattice data, we extend that analysis to construct a complete Lattice-based Equation of State (EoS) at finite density, which features a critical point at $(T, μ_B) \approx (114, 602)$ MeV. By integrating the extrapolated entropy density with respect to temperature, we reconstruct the pressure, baryon density, susceptibility, and speed of sound in the critical region, and analyze the focusing behavior of isentropic trajectories in the vicinity of the critical point.

We develop methods to obtain the fully differential cross-section for the $f \bar{f} \to Z(\ell\ell)\,h$ process to any desired order in effective field theory (EFT). To achieve this, we first derive a mapping between the partial wave expansion and the EFT expansion to all orders. We find that at lower orders, EFT predicts correlations between the different partial wave coefficients. This allows us to construct linear combinations of partial wave coefficients that get their leading contributions from a higher dimension EFT operator. We then introduce experimental observables, the so called angular moments -- that probe these linear combinations of partial wave coefficients -- and can be determined from a fully differential analysis of the angular distribution of the leptons arising from the $Z$ decay. We show that analysing the dependence of these angular moments on the $Zh$ invariant mass allows us to systematically probe all higher dimension EFT operators contributing to this process. While we take the Higgs-strahlung process as an example, the methods developed here are completely general and can be applied to other 2-to-2 collider processes.

This article presents an evaluation of off-the-shelf commercial accelerometers at the mixing chamber stage of a cryogen-free dilution refrigerator at temperatures down to 8 mK. In addition, we present results of radioassay of accelerometer samples using a high purity germanium detector counting setup. Cryogen-free dilution refrigerators using pulse-tube cryocoolers (PTs)-due to recent advances in their cooling capacity, long-term stability, and operational costs-have become ubiquitous tools in a wide range of fields ranging from experimental particle physics to quantum information sciences. However, vibrations induced by PTs can negatively impact the experimental payload in these applications. This work demonstrates that commercially available accelerometers can not only measure vibrations at millikelvin cryogenic temperatures but also pave the way for continuous, in situ, real-time vibration monitoring of dry dilution refrigerators.

A machine-learning-based framework for constructing generator-level observables optimized for parameter extraction in particle physics analyses is introduced, referred to as the Optimal Observable Machine (OOM). Unfoldable differential distributions are learned that maximize sensitivity to a parameter of interest while remaining robust against detector effects, systematic uncertainties, and biases introduced by the unfolding procedure. Detector response and systematic uncertainties are explicitly incorporated into the training through a likelihood-based loss function, enabling a direct optimization of the expected measurement precision while minimizing the bias from any assumption on the parameter of interest itself. The approach is demonstrated in an application to top quark physics, focusing on the measurement of a recently observed pseudoscalar excess at the top quark pair production threshold in dilepton final states. It is shown that a generator-level observable with enhanced sensitivity and long-term reinterpretability can be constructed using this method.

The exact localization result for the expectation value of the $1\over 2$ BPS circular Wilson loop in ${\cal N}=4$ SYM theory is given in the planar limit by the famous Bessel function expression: $\langle W\rangle = {2N\over \sqrt λ} I_1 ( \sqrt λ)$. Expanded in large $λ$ and expressed in terms of the AdS$_5 \times $S$^5$ string tension $T= {\sqrt λ\over 2π}$ this gives $\langle W\rangle = {\sqrt T\over 2πg_s} e^{2πT} (1- {3\over 16 π} T^{-1} + ...)$.The exponential is matched by the value of the action of the string with the AdS$_2$ world volume while the prefactor comes from the 1-loop GS string correction. Here we address the question of how the subleading $T^{-1}$ term could be reproduced by the 2-loop correction in the corresponding partition function of the AdS$_5 \times $S$^5$ GS string expanded near the AdS$_2$ minimal surface. We find that the string correction contains a non-zero UV logarithmic divergence implying that comparison with the SYM result requires a particular subtraction prescription. We discuss implications of this conclusion for checking the AdS/CFT duality at strong coupling.

We extend the Lorentzian replica framework for computing entanglement entropy to fully time-dependent gravitational settings, with emphasis on cosmological spacetimes without boundaries and gravitational theories lacking a dual quantum field theory description. Our approach constructs the replica path integral directly in real time, avoiding reliance on Euclidean continuation or time-reflection symmetry, and identifies the geometric conditions under which Lorentzian replica saddles appear in dynamical backgrounds. We analyze replica wormholes in both contexts, recovering a generalized version of the island rule. For completeness, we provide a concise review of the existing Lorentzian replica construction for quantum field theories and their holographic duals, which forms the foundation for our generalization.

The exploration of symmetry laws stands as a cutting-edge direction in modern physics research. This work delves into the examination of P and CP symmetry properties within the charm quark system by analyzing asymmetry parameters in the two-body decay process of $Ω_c$. By accounting for the polarization effects of electron and positron beams and employing the helicity formalism, we systematically analyze the decay characteristics of $Ω_c$ and its subsequent hyperon decays through specific asymmetry parameters. A comprehensive formulation of the angular distribution for these decay processes has been developed. The research assesses the detection sensitivity of asymmetry parameters in the $Ω_c\rightarrow Ω^-π^+$ decay mode across different experimental conditions, including varying data sample sizes and beam polarization configurations. These results contribute to enriching a theoretical foundation for forthcoming experimental endeavors at the STCF, offering significant implications for symmetry studies in the charm sector.

We present new constraints on the cosmic ray electron boosted light dark matter (CReDM) using the 205.4 kg$\cdot$day data of the CDEX-10 experiment located at the China Jinping Underground Laboratory. The cosmic ray electron spectrum and distribution in the Galaxy are generated by the $\tt GALPROP$ code package. In the calculation process of DM-electron scattering process in the Galaxy, we consider the energy-dependency of the DM-electron scattering cross section. The constraints on CReDM are set for both heavy and light mediator scenarios using the CDEX-10 dataset. The result exceeds previous Standard Halo Model (SHM) limits for DM mass lower than 0.6 MeV in heavy mediator case and corresponds to the best sensitivity among all direct detection experiments from 1 keV to 0.5 MeV in the light mediator scenario.

The deconfining transition in $SU(3)$ gauge theory, traditionally interpreted through the Gross-Witten-Wadia (GWW) model as a sharp third-order phase transition in the large-$N_c$ limit, appears as a smooth crossover in lattice QCD. This work demonstrates that the transition is topologically smoothed into a crossover by incorporating the fractal momentum space structure inherent to thermofractals. By matching the non-extensive $β$-function to one-loop QCD results, a fundamental scaling of the thermofractal index $q$ is derived as a function of the number of flavours $N_f$. It is proven that applying a $q$-deformed derivative operator $\mathcal{D}_q$ to the $q$-logarithm of the eigenvalue distance results in a non-extensive measure that effectively smears the topological stiffness of the gauge vacuum. A unified master equation for the Polyakov loop $\langle L \rangle$ is presented, governed by the thermofractal index $q$ and a single variance parameter $σ^2(T)$ that scales as $T^{1/(q-1)}$. The observed phase dynamics are shown to be asymptotic limits of this unified density: a ``soft'' algebraic growth $\langle L \rangle \propto T^{11}$ in the 1D string-like confined regime for $N_f=0$, and a rapid $1 - \langle L \rangle \propto T^{-21}$ suppression in the 3D deconfined volume for $N_f=3$. This approach provides a microscopic foundation for partial deconfinement theory and reproduces lattice QCD data with a reduced $χ^2 \approx 1.12$, offering a rigorous reconciliation between matrix model topology and the continuous QCD crossover.

The BESIII experiment has collected 2.6 billion $ψ(3686)$ events, 10 billion events, 20 $fb^{-1}$ of $D$ meson pairs at 3.773 GeV, and 7.33 $fb^{-1}$ of $D_sD_s^*$ events from 4.128 to 4.226 GeV. These huge data samples allow us to search for rare or forbidden processes in charm hadron decays. We summarize the recent research of charm rare decays at BESIII in this paper.

Detector visualization plays a vital role in high energy physics (HEP) experiments, yet existing detector descriptions, such as GDML, lack compatibility with industrial 3D tools. We present an automated conversion framework that transforms four major HEP detector descriptions, including GDML, Geant4, ROOT and DD4hep, into standardized FBX models compatible with a industrial 3D platform called Unity. This solution enables HEP detectors to be directly visualized in the professional 3D ecosystem, which is of great help for detector design verification, event display development, and public participation.

In arXiv:2211.09069, significant progress was made in decomposing simple products of fermion correlation functions, and in summing over spin structures of superstring amplitudes in genus two under cyclic constraints. In this manuscript we consider part of the same subject using a framework in which one of the branch points of the genus two curve is fixed at infinity. This framework is a direct generalization of the popular one in the case of genus one. We address some of the issues that remained unresolved in our previous paper arXiv:2209.14633. We show that the spin structures of the simple products of fermion correlation functions with cyclic conditions depend only on the Pe-function values at the half-periods of the genus two surface, for any number of factors in the products. Similar to the genus one case, we can provide basis functions to decompose the product. Consequently, the trilinear relations found in arXiv:2211.09069 can be derived from the known set of differential equations of genus two Pe-functions by simply setting the variables equal to the half-periods of the non-singular and even spin structures, as is the case for genus one. The focus of this manuscript is on the procedures for expressing the results of decomposed formulae in terms of the unique genus two theta function. At present we cannot provide a procedure for deriving the general form of the decomposed formula totally expressed in terms of the theta functions for an arbitrary number of the fermion correlation functions in the product, by the reason described in the text. We present some general results and demonstrate that concrete expressions of both the spin structure dependent and independent parts will be derived and simplified to analyze using the logic of the derivations of the classical solutions to Jacobi inversion problem and their modifications which will be given in this manuscript.

In this work, we perform the first numerical study of strong scalar backreaction in spectator chromo-natural inflation (SCNI) in the case where the spectator sector decays during inflation. The tachyonic instability in scalar fluctuations, activated as the system crosses the $m_Q = \sqrt{2}$ threshold, amplifies perturbations and may significantly alter the background dynamics. The strong scalar backreaction regime introduces an effective quartic term in the potential for the gauge field background that rapidly drives it to zero, accelerating the axion-gauge system decay. We describe the dynamics of such decay and derive the gravitational wave spectrum for a set of benchmark parameters. Interestingly, the signal may peak at interferometer scales and lie within LISA's projected sensitivity.

The $^8Be$ anomaly reported by the Atomki experiments can be explained by the hypothesis of an $X(17)$ boson that interacts with electrons via the ``Vector $\pm$ Axial-vector'' ($V \pm A$) interaction. Using existing experimental data, we derive constraints on the couplings of the $X(17)$ boson to electrons within this $V \pm A$ framework. With this setup, we attempt to identify $X(17)$ signals in the $e^+e^- \longrightarrow X(17) \longrightarrow e^+e^-$ process at the PADME experiment and in the $e^+e^- \longrightarrow X(17)γ\longrightarrow e^+e^- γ$ process at the BESIII experiment. Our findings indicate that observing the $X(17)$ signal at the PADME experiment is pessimistic, whereas the BESIII experiment may provide a definitive answer regarding the $X(17)$ hypothesis.

We reconsider the detailed structure of the topological character of the instantons in pure Yang-Mills theory on $S^1\times\mathbb{R}^3$, so-called calorons. The claim is that the standard formula for the topological character, the second Chern number, requires some modification through analytic consideration. For concreteness, we explicitly calculate the second Chern number of the gauge configuration of the Harrington-Shepard type with unit topological charge of the gauge group $\mathrm{SU}(2)$ in several gauges. The genuine formula is shown to be applicable even though the gauge connection is in singular gauge. The gauge dependence of the magnetic charge is also discussed.

In this Letter, we propose a strategy to extract information on the hierarchical amplitude pair in singly Cabibbo-suppressed (SCS) charmed baryon two-body decays, with a dominant amplitude proportional to $λ_s = V_{cs}^* V_{us}$ from tree operators and sub-leading one proportional to $λ_b = V_{cb}^* V_{ub}$ from both penguin and tree contributions. The coexistence of these two amplitudes is essential for generating nonzero CP violation (CPV) effects. Since the $λ_b$ amplitude is strongly suppressed, its experimental determination is highly challenging. However, by exploiting SU(3) flavor symmetry, which relates the well-measured Cabibbo-favored (CF) amplitudes to the SCS tree amplitudes, information on the $λ_b$ amplitude can be extracted.Using current experimental data, a conservative analysis yields $λ_b$ amplitudes can be as large as about $10\%$ of the corresponding tree amplitudes with a significance of $2.1σ$. In addition, the Lee-Yang parameters of these decays provide an independent probe of this elusive term. We further identify two golden decay channels, $Ξ_c^0 \to p K^-$ and $Ξ_c^0 \to Σ^+ π^-$, which are particularly well suited for experimental studies of CPV.

We present precision spectroscopy and isotope shift measurement of the ${}^1S_0 \leftrightarrow {}^3P_2$ clock transition in neutral ytterbium ($\mathrm{Yb}$) atoms. By revealing a magic wavelength at $905.4(2)$ nm, we successfully achieve the atomic spectrum narrower than $100$ Hz. The interleaved clock operation between isotopes allows us to determine isotope shifts of four bosonic isotope pairs at Hz-level uncertainties, which is combined with those of other four ultra-narrow transitions in $\mathrm{Yb}$ and $\mathrm{Yb}^+$ to construct the King plot. Importantly, the new isotope shift data reported in this work is a key to exclude the possibility of attributing the observed nonlinearity of the three-dimensional King plot solely to the new physics, while the previous works rely on the other terrestrial bound set by the neutron scattering and $(g-2)_e$ measurements. This work paves the way for the effective use of precision isotope shift data in the King plot analysis and stimulates further measurements in $\mathrm{Yb}$ and other elements.

We derive a parameter-free geometric constraint on residue phases dictated by the pole-threshold angle. Using the N(2190) anomaly as a test case, this constraint reveals a sign ambiguity in prior data; correcting it yields a phase of $-28^\circ\pm10^\circ$, matching our prediction. This consistency validates the method as a model-independent diagnostic for distinguishing compact from molecular states, offering a rigorous tool for exotic spectroscopy.

When studying the Casimir effect in a quantum field theory setting, one can impose the boundary conditions by adding appropriate Dirac-delta functions to the path integral. In this paper, the limits of this approach are explored under different boundary conditions.

Recently, the LHCb collaboration reported the first observation of $CP$ violation in baryon decays, with a significance of more than $5σ$. This strongly motivates us to investigate the $CP$ violation in more baryon decay processes. In this work, we employ the final-state rescattering mechanism with introducing two model parameters, $Λ_{charm}$ and $Λ_{charmless}$, and calculate two-body non-leptonic baryon decays $Λ^0_b \rightarrow Λ(1520)\,π^0/κ(700)/f_0(500, 980)/ρ^0/K^{*0}/φ$ and $Ξ^-_b \rightarrow Λ(1520)\,K^-$. Consequently, we evaluate the corresponding branching ratios, $CP$ asymmetries, and interference effects between different decay amplitudes. Our theoretical predictions for certain decay channels are in good agreement with current experimental measurements, while the remaining processes--particularly the remarkably large $CP$ violation observable revealed by the kinematic analysis are expected to be tested in future experiments.

In this {\it Letter}, we present a compelling and robust argument for the roles of neutrino fast flavor conversion (FFC) in the explosion mechanism of core-collapse supernova (CCSN), combining the {\it multi-angle} FFC subgrid model rooted in quantum kinetic theory with the multi-dimensional four-species Boltzmann neutrino radiation hydrodynamics. Employing various progenitor masses and the nuclear equations of states, we find that the effect of FFC on CCSN explosion is bifurcated depending on the progenitors. For the lowest-mass progenitor, FFC facilitates the shock revival and enhances the explosion energy, whereas for higher-mass progenitors its impact is inhibitory. We identify the mass accretion rate as the key determinant governing this bifurcation. When the mass accretion rate is low (high), the contribution of FFC to neutrino heating becomes positive (negative), because the heating efficiency enhancement via FFC-driven spectral hardening of electron-type neutrinos dominates over (is outweighed by) the concurrent reduction in neutrino luminosity. Our results further highlight the limitations of approximate neutrino transport, and demonstrate that a multi-angle treatment is essential for accurately capturing FFC effects; otherwise, FFCs are missed and even generated spuriously.

Axions coupled to thermal non-Abelian gauge fields may have cosmological significance. As the heat bath defines a frame, its influence depends separately on energy and momentum. A light-like momentum ($k \approx ω$) is relevant for the axion contribution to the effective number of light neutrinos, $ΔN^{ }_\mathrm{eff}$, whereas a vanishing momentum ($k=0$) plays a role for warm natural inflation or ultralight dark matter, and has been employed in lattice estimates (both classical and quantum-statistical) of the strong sphaleron rate. Focussing on soft energies ($α_\mathrm{s}^{ }T \ll ω\ll πT$), we carry out an HTL computation to show how the domains $k=0$ and $k \approx ω$ interpolate to each other. We then compare with lattice data at $k=0$, and connect our analysis to NLO computations at $k \approx ω\ge πT$. Assembling the current best input, we re-investigate light QCD axion decoupling dynamics at $T \ge 200$ MeV, showing that efficient interactions in the ultrasoft domain increase $ΔN^{ }_\mathrm{eff}$ from $\sim 0.03$ to $\sim 0.04$ at $f^{ }_a = 4\times 10^8_{ }$ GeV.

The BESIII experiment at BEPCII provides a clean environment to test baryon-dark sector connections and baryon-number violation. Using $10^{10}$ $J/ψ$ events, searches were performed for $Σ^{+}\to p+\text{inv}$, $Ξ^{-}\toπ^{-}+\text{inv}$, and $\barΛ-Λ$ oscillations. No excess was observed, yielding $\mathcal{B}(Σ^{+}\to p+\text{inv})<3.2\times10^{-5}$, $\mathcal{B}(Ξ^{-}\toπ^{-}+\text{inv})<(4-7)\times10^{-5}$, and $P(Λ)<4.4\times10^{-6}$ (90\% C.L.). These are the most stringent limits on invisible baryon decays and $ΔB=2$ transitions in strange baryons, probing new physics beyond the Standard Model.

Recently, the exciting new Fermilab (FNAL) Muon g-2 measurement impressively confirmed the final Brookhaven (BNL) result from 2004, and with a result four times more precise, has launched a new serious attack on the Standard Model (SM). On the theoretical side, ab initio lattice QCD (LQCD) calculations of hadronic vacuum polarization have made remarkable progress. They are now the new standard for studying the leading non-perturbative contributions, which have previously hindered matching with the precision required for full exploitation of the experimental results. The lattice results affected both leading hadronic contributions the hadronic vacuum polarization (HVP) and the hadronic light-by-light (HLbL) contributions by increasing the previously generally accepted $e^+e^-$ to hadrons based dispersion relation results. The shifts reduced the discrepancy between theory and experiment, leaving nothing missing. One of the most prominent signs of Beyond the Standard Model (BSM) physics has disappeared: the SM appears validated more than ever, in agreement with what other searches at the Large Hadron Collider (LHC) at CERN tell us! A triumph of the SM, even though the SM cannot explain known cosmological puzzles like dark matter or baryogenesis, and why neutrino masses are so tiny, the absence of strong CP violation, for example. I also argue that the discrepancy between the data-driven dispersive result and the lattice QCD results for the hadronic vacuum polarization can be largely explained by correcting the $e^+e^-$ data for 'rho-gamma' mixing effects.

The Light Dark Matter eXperiment (LDMX) is a proposed fixed-target missing momentum search for sub-GeV thermal relic dark matter. LDMX aims to probe thermal dark matter targets with 1016 electrons on target. Such an approach requires a high-repetition rate, low-current beam, with an average of one electron on target per event. These requirements are well-suited to the DArk Sector Experiments at LCLS-II (DASEL) facility, which will take advantage of the unused RF buckets between LCLS-II bunches to produce a well-defined low-current beam with a 26.9 ns bunch spacing. This document describes the results of a measurement of dark current in the Sector 30 transfer line (S30XL) of the LCLS-II beam, using a prototype of the LDMX trigger scintillator (TS) subsystem.

Massive stars are the key engines of the Universe. However, their evolution and thus their ionizing feedback are still not fully understood. One of the largest gaps in current stellar evolution calculations is the lack of a model for the mass ejections that occur when the stars reach the Eddington limit, such as during an Luminous Blue Variable (LBV) phase. We aim to remedy this situation by providing a physically motivated and empirically calibrated method applicable in any 1D stellar evolution code to approximate the effect of such mass loss on stellar evolution. We employ the 1D stellar evolution code MESA, in which we implement a new mass-loss prescription that is acting when stellar models inflate too much when reaching the Eddington limit. Synthetic massive-star stellar populations using calculated grids of single-star models with this mass loss prescription are compared with the observed populations in the Large and Small Magellanic Clouds. In combination with already computed grids of binary evolution models, we investigate the impact of binarity on our predictions. Our single-star models reproduce key features of the observed stellar populations, namely (i) the absence of stars located beyond the Humphreys-Davidson limit, (ii) an upper limit of RSG luminosities, (iii) the faintest observed single WR stars, (iv) the absolute number of O-stars, WRs, and RSGs, (v) WO stars in low metallicity environments, and (vi) the positions of LBV stars in the HRD. Our binary population explains at the same time the 70% binary fraction of O-stars and the 40% binary fraction of WR stars. However, our synthetic population also has caveats, such as an overproduction of bright H-free WN stars. Our results show that the effect of Eddington-limit induced mass ejections on the structure and evolution of massive stars can remove tension between predicted and observed massive star populations.

To reproduce observed galaxy properties, cosmological simulations require that massive galaxies experience feedback from active galactic nuclei, which regulates star formation within those galaxies. However, the energetics and timescales of these feedback processes are poorly constrained. We combine optical, infrared, sub-millimeter and radio observations of the active galaxy VV 340a, hosting a low-power jet launched from a supermassive black hole at its center. We find that the jet undergoes precession, with a period of (8.2 $\pm~$5.5) $\times~$10$^5$ years, and drives an outflow of gas at a rate of 19.4 $\pm~$7.9 solar masses per year. The jet shocks the gas, producing highly ionized plasma extending several kiloparsecs from the nucleus. The outflow ejects sufficient gas from the galaxy to influence its star formation rate.

Radio mini-halos are perplexing features, typically hosted by X-ray cool-core galaxy clusters. Understanding the connection between thermal X-ray and non-thermal radio emission is key to uncovering their origin. Here, we present a multiwavelength study of the cool-core cluster RXCJ1558.3-1410 using archival Chandra X-ray and wideband uGMRT radio data (Bands 3, 4 and 5). Our improved analysis confirms a previously known X-ray cavity at $\sim$36 kpc south-east of the cluster centre and we report a new cavity at $\sim$42 kpc to the north-west. These cavities suggest that the AGN provides mechanical power of $\sim$$6.0 \times 10^{44}$ erg s$^{-1}$, sufficient to offset radiative cooling in the ICM. We also detect a sharp surface brightness edge at $\sim$72 kpc south-east of the centre, characterised by a temperature jump and pressure continuity, consistent with a cold front, likely caused by gas sloshing from a minor merger. Our uGMRT images reveals an interesting diffuse emission surrounding the brightest cluster galaxy (BCG), with its edge spatially coinciding with the sloshing cold front and roughly with the cooling radius. Furthermore, a low star formation rate and uniform metal abundance up to the sloshing edge are consistent with the earlier findings of suppression of star formation and metallicity homogenisation by mixing core gas through sloshing. Finally, the spatial correlation between the mini-halo and the observed X-ray features indicates that ICM sloshing, rather than AGN feedback, plays a dominant role in powering the proposed radio mini-halo emission.

We present new precise measurements of electron temperatures and oxygen abundances in the southeast knot of I Zw 18, one of the most metal-poor blue compact dwarf galaxies known, using spectroscopic data from the Dark Energy Spectroscopic Instrument Data Release 1 (DESI DR1). For the first time in I Zw 18, we directly measure electron temperature in the low-ionization zone using the rarely detected [O II] $λ\lambda7320,7330$ doublet. We also detect the auroral lines [O III] $\lambda4363$ and [S III] $\lambda6312$, associated with high and intermediate ionization zones, respectively. We derive $T_{\mathrm{e}}([\mathrm{O}\,\mathrm{III}])=21\,200\pm860\ \mathrm{K}$, $T_{\mathrm{e}}([\mathrm{O}\,\mathrm{II}])=16\,170\pm950\ \mathrm{K}$, and $T_{\mathrm{e}}([\mathrm{S}\,\mathrm{III}])=17\,290\pm1750\ \mathrm{K}$, highlighting a significant temperature difference between ionization zones. Using these direct temperature measurements, we determine a total oxygen abundance of $12+\log(\mathrm{O}/\mathrm{H})=7.066\pm0.046$, $\log(\mathrm{N}/\mathrm{O})=-1.509\pm0.097$, and $\log(\mathrm{S}/\mathrm{O})=-1.558\pm0.041$. Our results extend the calibration of $t_2$--$t_3$ relations to the highest temperatures, providing important anchor points for the temperature structure of extremely metal-poor H II regions, including high-redshift galaxies where direct temperature measurements are especially challenging.

Relativistic jets launched in binary neutron star (BNS) mergers are widely accepted as the engines powering most of the population of short gamma-ray bursts (GRBs). Understanding their structure and dynamics-particularly during and after breakout from the merger ejecta-is crucial for interpreting GRB afterglows, especially for off-axis observers. Traditional models often assume simple angular or radial jet profiles, potentially missing key features emerging for jets piercing through realistic environments. This work aims to investigate the formation and evolution of the jet structure as it propagates through a non-homogeneous, anisotropic BNS merger environment. We focus on how the interaction with the ambient medium shapes the jet's angular and velocity distributions and assess the impact of this realistic structure on the resulting afterglow light curves. We perform a series of 3D relativistic magnetohydrodynamic simulations of jets launched in post-merger environments, exploring different injection conditions. Simulations are evolved to late times, approaching the ballistic regime, where further dynamical evolution becomes negligible. From the resulting outflows, we extract energy and velocity profiles and compute multi-wavelength afterglow light curves using a semi-analytic model that includes radial stratification and the full 3D jet geometry. More energetic or earlier-launched jets drill more efficiently through the ejecta, but all develop asymmetries that leave clear imprints in the off-axis afterglow light curves. All models exhibit a complex multi-shock breakout structure responsible for an early, dimmer peak in the afterglow. Despite structural differences, all simulated jets are consistent with the observational data of the multi-messenger BNS merger event GW170817.

JWST observations have suggested that some high-redshift galaxies may be ultra-massive, thereby challenging standard models of early galaxy formation and cosmology. We analyse the stellar masses using different modelling assumptions and with new data of three galaxies (S1, S2 and S3), whose NIRCam/grism redshifts were consistent with $z>5$. These three optically dark galaxies have previously been reported to host exceptionally high stellar masses and star-formation rates, implying extremely high star-formation efficiencies. Recent NIRSpec/IFU observations for S1 indicate a spectroscopic redshift of $z_{\rm spec}=3.2461^{+0.0001}_{-0.0002}$, which is lower than previously reported. Using the Bayesian spectral energy distribution (SED) modelling tool \texttt{Prospector}, we investigate the impact of key model assumptions on stellar mass estimates, such as the choice of star-formation history (SFH) priors (constant versus rising SFH base for the non-parametric prior), the dust attenuation law, and the treatment of emission line fluxes. Our analysis yields revised stellar masses of $\log(M_{\star}/M_{\odot}) \approx 10.36^{+0.47}_{-0.32}, 10.95^{+0.11}_{-0.10}$ and $10.31^{+0.24}_{-0.19}$ for S1, S2, and S3, respectively. We find that adopting a rising SFH base prior results in lower inferred stellar masses compared to a constant SFH base prior. We identify a significant degeneracy between the dust attenuation curve slope, the amount of dust attenuation, and stellar mass. Our results highlight various systematics in SED modelling due to SFH priors and dust attenuation that can influence stellar mass estimates of heavily dust obscured sources. Nevertheless, even with these revised stellar mass estimates, two of the three galaxies remain among the most massive and actively star-forming systems at their respective redshifts, implying high star-formation efficiencies.

We present new results from the MUSE-ALMA Haloes survey, covering 79 galaxies associated with strong HI absorption at redshift about 0.5. Our ALMA Cycle 10 observations add 39 systems to the initial 21, bringing the total to 60 galaxies. CO emission is detected in 9 new galaxies, and in 12 of 60 total, doubling the number of CO-emitting HI-selected galaxies and probing 1.2 dex deeper in molecular gas mass than previous studies. These galaxies span a wide range of stellar masses and metallicities. By comparing CO(2-1) and CO(3-2) properties with star formation rates and gas-phase metallicities from VLT/MUSE and HST, we find a dual behaviour in star formation efficiency: low-mass systems follow main-sequence scaling relations, while high-mass systems show suppressed star formation. This diversity indicates that HI absorbers trace both evolved and younger galaxies, providing a key step toward completing the baryon census at redshift about 0.5.

We point out that on 22 January 2026, the interstellar object 3I/ATLAS will align to within an exceptionally small angle, alpha= 0.69 degrees, with the Earth-Sun axis. This rare alignment provides unique circumstances for measuring the opposition surge and polarimetric properties of interstellar cometary dust. We characterize the alignment geometry, outline key scientific opportunities, and define the observational requirements for data collection. Observations before and after the alignment time offer an unprecedented opportunity which may not repeat for decades, for characterizing the albedo, structure, and composition of interstellar matter.

We present independent polarimetric observations of the interstellar object 3I/ATLAS, including the first near-infrared polarimetric measurements. Using imaging polarimeters, we measured the degree of linear polarization from the visible RC band (0.64 μm) to the near-infrared KS band (2.25 μm), and investigated its dependence on solar phase angle (polarization phase curve; PPC) and wavelength (polarization color curve; PCC). We confirm that the PPC of 3I/ATLAS differs significantly from those of typical Solar System comets, showing an unusually large polarization amplitude. This PPC shows no significant change in the RC band across perihelion passage, despite the perihelion lying within the water snow line. This indicates that the unusual polarimetric behavior of 3I/ATLAS is unlikely to be driven by transient volatile activity, but instead reflects intrinsic optical properties of refractory dust particles. The PCC increases with wavelength over 0.6-1.2 μm and peaks at 1.5-2.0 μm, suggesting that the dominant scattering units are dust aggregates composed of submicron-sized monomers, broadly consistent with interstellar dust and solar-system cometary aggregates. Taken together, our results indicate that 3I/ATLAS preserves polarimetric properties characteristic of a primitive cometary planetesimal formed in another planetary system, with a refractory dust composition that differs from that typically observed among Solar System comets, despite sharing a similar size scale of the aggregate building blocks.

The 2175 Angstrom extinction bump, the most prominent spectral feature superimposed on the interstellar extinction curve, is widely seen in the interstellar medium (ISM) of the Milky Way and external galaxies, both near and far. While its central wavelength is remarkably stable and independent with environment, its width shows considerable variation and environmental dependence. Here we examine the extinction curve for the line of sight toward HD 93222, a young star located in the Carina nebula. It is found that the 2175 Angstrom bump is extremely sharp, which is among the narrowest ever found in the Milky Way and external galaxies. We model the derived extinction curve and find that, to explain the extinction characteristics of HD 93222, in addition to the conventional silicate and graphite dust mixture, an additional population of nano-sized graphitic grains is required.

The initial mass function (IMF) is a cornerstone of star formation studies, yet its universality remains debated. We investigate the IMF in the young massive cluster RCW 36, located in the Vela Molecular Ridge and comparable to the Orion Nebula Cluster in stellar density. Our goal is to build the most complete census of RCW 36 and derive its first IMF and star-to-brown-dwarf (BD) ratio. We combine new GLAO observations from HAWK-I/VLT with archival data (2MASS, SOFI/NTT) and Gaia DR3 kinematics. Photometric accuracy and source extraction were improved using \textsc{DeNeb}, a deep-learning algorithm that removes complex nebular emission. Membership probabilities were assigned via color-magnitude diagram comparisons with a control field, and stellar masses were estimated using model isochrones. We find a revised distance of $954\pm40\,$pc and determine the IMF down to $\sim0.03\,M_{\odot}$, described by a broken power law ($dN/dM\propto M^{-α}$) with $α=1.62\pm0.03$ for $0.20$-$20\,M_{\odot}$ and $α=0.46\pm0.14$ for $0.03$-$0.20\,M_{\odot}$. The star-BD ratio is $2$-$5$, consistent with other Galactic clusters. Lastly, through a study of the differences in the IMF within and outside $0.2\,$pc and the cumulative mass distributions for low-mass and intermediate to high-mass sources, we also detected signs of possible mass segregation within RCW 36, which should be primordial. RCW 36 shares many characteristics with other young massive clusters, such as a shallower than Salpeter high-mass slope and the possibility of mass segregation. The flatter lower-mass regime of the IMF is similar to most Galactic clusters. The star-BD ratio is also in line with the observed values in other clusters, independent of their inherent properties.

Gamma ray bursts (GRBs) offer a powerful probe of the cosmic expansion history far beyond the redshift range accessible to Type Ia supernovae. However, the calibration of GRB luminosity correlations is hindered by the circularity problem, which arises from assuming a fiducial cosmological model during calibration. In this work, we perform a model independent calibration of GRB luminosity relations using observational Hubble parameter H(z) data from the A220 and J220 compilations, thereby avoiding explicit cosmological assumptions. We employ Artificial Neural Network (ANN) to reconstruct the calibration relation directly from the data. In addition, we implement a Bayesian Neural Network (BNN) framework as an alternative approach, enabling a data driven treatment of both statistical and systematic uncertainties. The calibrated GRB sample is used to constrain the Amati relation, and we systematically compare the outcomes obtained from different calibration techniques and datasets. While the Amati Parameters obtained from GRBs caibrated from the ANN and BNN results are consistent with previous low redshifts calibrations using model-independent methods, the BNN approach provides a more robust framework.

PSR J1455$-$3330 is a $\sim$7.98 ms pulsar in a $\sim$76.17 day nearly circular orbit with a white dwarf companion. In this work, we combine the available Lovell, Nançay decimetric Radio Telescope, Green Bank, and MeerKAT pulsar timing data spanning $\sim$ 30 years to measure the kinematic and relativistic effects of PSR J1455$-$3330 to constrain its 3D orbital geometry and component masses. We detect a relativistic Shapiro delay signal. We measure a significant orthometric amplitude $h_3 = 0.307^{+0.022}_{-0.026}$ $μ$s and an orthometric ratio $ς= 0.551^{+0.057}_{-0.054}$. We measure the change in projected semi-major axis $\dot{x} = -202.1^{+2.5}_{-2.7} \times10^{-16} \, \rm s\,s^{-1}$ with high significance, parallax, $\varpi$ = 1.11(6) mas, parallax derived distance 0.90(5) kpc, and a precise total proper motion magnitude of 12.432(2) mas yr$^{-1}$. A self-consistent analysis of all kinematic and relativistic effects, assuming general relativity, yields two solutions: (1) a pulsar mass $M_{\rm p} = 1.39^{+0.38}_{-0.18}\, \rm M_{\odot}$, a companion mass $M_{\rm c} = 0.293^{+0.056}_{-0.026}$ $\rm M_{\odot}$, an orbital inclination, $i = 63(2)^{\circ}$, and longitude of the ascending node, $Ω= 212(12)^{\circ}$ or (2) a pulsar mass $M_{\rm p} = 1.53^{+1.10}_{-0.22} \, \rm M_{\odot}$, a companion mass $M_{\rm c} = 0.309^{+0.163}_{-0.026}\, \rm M_{\odot}$, an orbital inclination, $i = 123(4)^{\circ}$, and longitude of the ascending node, $Ω= 334(12)^{\circ}$. All uncertainties represent the 68.27$\%$ credibility region. These results strongly favour a helium-dominated white dwarf companion.

We revisit the Great Attractor using the Manticore-Local suite of digital twins of the nearby Universe. The Great Attractor concept has been proposed as an answer to three distinct questions: what sources the Local Group velocity in the cosmic microwave background frame, where present-day velocity streamlines converge, and where the Local Group is moving to. Addressing the original motivation of the Great Attractor -- explaining the Local Group cosmic velocity -- we find that mass within $155~h^{-1}\mathrm{Mpc}$ accounts for only ${\sim}72\%$ of that velocity magnitude with ${\sim}38\,°$ directional offset. We show that even in the purely linear regime convergence within this volume is not guaranteed, particularly when also accounting for small-scale contributions to the observer velocity; no single structure, including the proposed Great Attractor, would be expected to dominate the velocity budget. Streamline convergence is smoothing-scale-dependent, transitioning from Virgo at small scales through the Hydra--Centaurus region at intermediate scales to Shapley at large scales; at intermediate smoothing the convergence point lies near Abell 3565 with an asymmetric basin of mass $\log( M / (h^{-1} \mathrm{M}_\odot)) = 16.4 \pm 0.1$ that excludes Norma. To address the third question, we evolve the Manticore-Local realisations to scale factor $a = 10$ in a new Beyond-Present-Time simulation suite and identify the asymptotic future location of the Local Group. We find that the dominant motion is towards Virgo, but even it contributes at most one third of the Local Group velocity. Our results demonstrate that the classical Great Attractor is not a dynamically dominant structure but an artifact of the instantaneous velocity field, and that no single attractor is likely to account for the Local Group motion in the cosmic rest frame.

Our Local Group, dominated in mass by the Milky Way (MW) and M31, provides a unique laboratory for testing $Λ$CDM cosmology on small scales owing to its proximity. However, its connection to the surrounding large-scale environment, which is essential for interpreting its properties, is inadequately understood. In this work, we explore the connection between Local Group analogues (LGAs) and their surrounding large-scale environments using the ABACUSSUMMIT simulation suite, highlighting the key role of the coupling energy of the MW-M31 orbit, $E_{\rm coupling}$. We find that LGAs with high $E_{\rm coupling}$ preferentially reside in denser regions, whereas those with low $E_{\rm coupling}$ tend to occupy low-density environments. Furthermore, LGAs with low $E_{\rm coupling}$ exhibit strong alignment with cosmic filaments, manifested as a pronounced polar anisotropy in the distribution of tracer haloes. By contrast, LGAs with high $E_{\rm coupling}$ show a weaker polar anisotropy but an enhanced azimuthal anisotropy, with large-scale tracer haloes preferentially lying in the plane spanned by the halo pair and the orbital spin vector. Within this framework, our Local Group is characterised by typical $E_{\rm coupling}$ residing in a relatively under-dense environment, yet it remains consistent with the 95\% range of analogue systems identified in the simulation.

The parameters of an axisymmetric model for the gravitational potential of the Galaxy have been refined. The basic curve of the Galaxy's rotation in a distance interval of $R:0-190$ kpc was constructed using the velocities of masers, classical Cepheids, Red Clump stars, Blue Horizontal Branch stars, halo stars, globular clusters, and dwarf satellite galaxies of the Milky Way. The rotation curve was selected in such a way that there would be no dominant burst of circular velocities in the central ($R<2$ kpc) region of the Galaxy. As a result, we constructed two two-component models of the galactic potential, which include contributions from the disk and the halo of invisible matter, as well as a three-component model with a small-mass bulge added in advance. These models can be useful in studying the long-term orbital evolution of stars and open and globular star clusters in the central ($R<4$ kpc) region of the Galaxy. The constructed models were tested for self-consistency by comparing their rotation curves with a set of model curves generated with the Illustris TNG50 software package.

Many stellar systems exhibit a finite spatial extent, yet constructing self-consistent spherical models with a prescribed outer boundary is non-trivial because sharp density cutoffs introduce discontinuities that lead to inconsistencies in the associated distribution function. In this paper we show that these difficulties arise from the abruptness of the truncation rather than from the finite extent itself. We introduce a general and infinitely differentiable radial truncation scheme that can be applied to any density profile, and illustrate its behaviour using the Hernquist model. We find that softly truncated models are dynamically consistent provided that the truncation is sufficiently gradual, and we determine the corresponding critical truncation sharpness. Their distribution functions display a characteristic bump-dip feature near the truncation energy that signals the transition between consistent and inconsistent cases. In contrast to sharply truncated models, softly truncated systems can support an extensive family of Osipkov-Merritt orbital structures, including moderately radial ones. Soft truncations therefore offer a general and physically motivated route to constructing finite-extent dynamical models with well-controlled outer-edge behaviour.

Rotation curves are a fundamental tool in the study of galaxies across cosmic time, and with the advent of large integral field unit (IFU) kinematic surveys there is an increasing need for efficient and flexible modelling tools. We present RotCurves, a parametric forward-modeling tool designed for rotation curve analysis at high-z, correcting for ``beam smearing" by projecting and convolving the beam PSF in the plane of the galaxy. We benchmark RotCurves against the established parametric code dysmalpy using synthetic observations. The typical runtime with RotCurves is a few ~10ms, a factor 250 faster than dysmalpy for a single realization. For well-resolved systems (PSF FWHM < Reff), the mock observed rotation and dispersion curves agree to within 5% up to 3Reff, where most of the discrepancies are in the inner disk. whereas in marginally resolved systems (PSF FWHM > 1.5 Reff) discrepancies increase to up to 15%. Using a built-in MCMC fitting procedure, RotCurves recovers well the intrinsic model parameters across a wide range of galaxy properties and accounting for realistic noise patterns. Systematic biases emerge for the effective radius and for low disk masses (Mdisk < 3x10^9 Msun). We show excellent parameter recovery at high signal-to-noise ratios (S/N > 25), with increasing deviations in parameter recovery at lower S/N. RotCurves is best suited for inclinations of 10 < i < 80. RotCurves is built as an exploratory tool for rapid testing of mass model assumptions, parameter studies and for efficiently processing large samples of observational data from large IFU surveys. The code is publicly available on github.

Aims. The blazars Mrk 421 and Mrk 501 have shown multiwavelength variability on all observed timescales, and have been well studied at high energies on short timescales. We aim to characterise the long-term temporal behaviour of these blazars at synchrotron energies, namely optical, UV, and X-ray, in order to assess current models of these objects and their processes. Methods. Amongst the longest light curves ever studied for these sources, we investigated 20 years of data (2005-2025) from the Swift-UVOT and Swift-XRT telescopes. We examined spectral models, fractional variabilities, flux distributions, and X-ray photon index vs flux relations, as well as carrying out in-depth time series analysis using structure functions, Lomb-Scargle periodograms, and discrete correlation functions. Results. Mrk 421 and Mrk 501 both showed intriguing variability in all studied wavelengths; this variability has been found to be energy dependent, as has the trend of lognormality in flux distributions. X-ray photon indices fluctuated greatly throughout the entire period, showing an overall harder-when-brighter trend. Hints of a quasi-periodicity have been found in the X-ray of Mrk 501 (host frame time scale $\sim390$ days, >3$σ$) but not in the UV or X-ray of Mrk 421, or in the UV of Mrk 501. No correlation at any time lag was found between the optical/UV and X-ray bands in either source.

According to the classical AGN model, broad emission lines originate from the broad-line region (BLR) and are observable only when the attenuation by the dusty torus is small. However, we recently found several heavily-obscured ($A_V > 50$ mag) AGNs with broad H$α$ detections: MCG -3-34-64, UGC 5101, and Mrk 268. To investigate the origin of the observed broad line in these AGNs, we performed multi-epoch optical spectroscopic observations to search for flux variability of the broad H$α$ line. For MCG -3-34-64 and UGC 5101, no significant variability was detected, suggesting that the broad line of these AGNs may arise from sources other than the BLR. Spectral fitting analysis suggests possible large contribution of ionized outflows to the observed broad component of MCG -3-34-64, while both the outflow and scattering by polar material can explain that of UGC 5101. For Mrk 268, we detected a significant ($4.3σ$) flux variation of the broad H$α$ line by using the flux ratio of the H$α$ complex and the [SII]$λ\lambda6716$, 6731 doublet, indicating that the broad line originates directly from the BLR. The lack of significant flux variation in the optical continuum implies that the line of sight to the nucleus of Mrk 268 is mildly obscured. Our results demonstrate that the observed broad H$α$ lines in obscured AGNs likely have multiple origins. Such complexity may introduce additional uncertainties in black hole mass measurements of distant AGNs revealed by e.g., JWST.

Three-nucleon force (3NF) plays an important role in understanding the structure of finite nuclei and the saturation properties of infinite nuclear matter. The chiral 3NF derived from the chiral effective field theory has been successful in $\textit{ab initio}$ studies of atomic nuclei. However, challenges remain, such as parameterizing low-energy constants and applying regulators. Most of established chiral nuclear forces have a nonlocal form in the momentum space. In this work, we construct local and hybrid local-nonlocal chiral 3NFs for the newly established Idaho local position-space two-nucleon potential, and calculate binding energies and radii of nuclei up to $^{132}$Sn. The two low-energy constants of 3NF are constrained by the ground-state energies of $^3$H and $^{16}$O, as suggested in a recent work. The chiral Hamiltonian obtained with the local-nonlocal regulator can simultaneously reproduce the experimental ground-state energies and charge radii of nuclei over a large range from $^4$He to $^{132}$Sn.

The Andes Large area PArticle detector for Cosmic ray and Astronomy (ALPACA) is a new air-shower array experiment under construction in the Bolivian Andes, and its prototype ALPAQUITA surface array has been operating since 2023 April. In addition to the traditional $\ge$3-hit or $\ge$4-hit coincidences to trigger recording air-shower events, ALPAQUITA records the counting rates of the $\ge$1-hit and $\ge$2-hit events (Any1 and Any2, respectively). We report a successful detection of a Forbush decrease occurred on 2024 May 10 caused by a passage of an interplanetary shock formed ahead of the Interplanetary Coronal Mass Ejection. The amplitude detected in the Any1 rate is 4.26$\pm$0.33% at the median primary rigidity of 76GV which is consistent with the observations with the worldwide neutron monitor and muon detector networks. Under the assumption of a power-law rigidity spectrum, we renormalized the errors of the observed amplitude ($A_{obs}$) and fitted them as a function of the median primary rigidity ($R_{m}$) of each detector and observational method. The result $A_{obs} = (10.9\% \pm 0.9\%) \times (R_{m}/10\,GV)^{-0.55 \pm 0.07}$ exhibits a hard nature of this event. Our non-detection in the Any2 rate decrease constrains the amplitude with a 2$σ$ upper limit to be 0.95% at 960GV. This marginally suggests an existence of a spectral softening between 100GV and 1000GV as also suggested by the Misato underground muon detector at 145GV. Although a strong geomagnetic storm was observed during this period, we conclude it does not impact our results. Our novel technique realizes a unique coverage to study the behavior of the Forbush decreases at the highest rigidity.

We compute the cosmic stellar, dust and neutral gas mass history at $0<z\lesssim3$ using ProSpect spectral energy distribution modelling of $\approx 800 \, 000$ galaxies in the Galaxy and Mass Assembly (GAMA) survey and the Deep Extragalactic VIsible Legacy Survey (DEVILS). The cosmic dust mass history broadly follows the shape of the cosmic star formation history; though, the decline is slower, suggestive of a slowing rate of dust growth and destruction as the star formation declines past its peak at $z\approx 2$. Neutral gas masses were estimated by scaling the dust masses by the metallicity-dependent dust-to-gas ratio. The neutral gas mass density as traced by the dust is an average of $\approx 0.6$ dex lower than that measured from $21$cm experiments, most likely due to differences in the spatial scales inhabited by dust and HI. Folding in measurements of the supermassive black hole mass density obtained previously with similar data and methods, we present a self-consistent census of the baryons confined to galaxies. Stars, neutral gas, SMBHs and dust contained within the optical radii of galaxies account for $\approx 5$ per cent of the baryons. Most of the remaining $\approx 95$ per cent of baryons must be ionised and dispersed throughout the interstellar, circumgalactic and intergalactic media within, around and between galaxies.

We present a measurement of Baryon Acoustic Oscillations (BAO) in the cross-correlation of triply ionized carbon C IV absorption with the positions of quasars (QSO) and Emission Line Galaxies (ELG). We use quasars and ELGs from the second data release (DR2) of the Dark Energy Spectroscopic Instrument (DESI) survey. Our data sample consists of 2.5 million quasars, 3.1 million ELGs, and the C IV absorption is measured along the line of sight of 1.5 million high redshift quasars with $z > 1.3$. We measure the isotropic BAO signal at 4.2$σ$ for the CIV$\times$QSO cross-correlation. This translates into a 3.0% precision measurement of the ratio of the isotropic distance scale, $D_{\rm V}$, and the sound horizon at the drag epoch, $r_{\rm d}$, with $D_{\rm V}/r_{\rm d}(z_{\rm eff} = 1.92) = 30.3 \pm 0.9$. We make the first detection of the BAO feature in the CIV$\times$ELG cross-correlation at a significance of 2.5$σ$ and find $D_{\rm V}/r_{\rm d}(z_{\rm eff} = 1.47) = 24.6 \pm 1.0$.

On the framework of relativistic force-free magnetohydrodynamic (MHD) turbulence, we explore the fundamental properties of strong and weak turbulent cascades using high-resolution numerical simulations in the presence of a uniform background magnetic field. We find that (1) power spectra and scale-dependent anisotropies both for the strong and weak turbulence resemble those observed in the non-relativistic MHD turbulence; (2) intermittency of magnetic fields in strong turbulence is stronger than that in the weak one; (3) generated Alfvén modes show similar energy spectra and scale-dependent anisotropies to those of non-relativistic case; (4) generated fast modes present a power spectrum similar to that of Alfvén modes, with a strong (for strong turbulence) or weak (for weak turbulence) scale-dependent anisotropy, which are significantly different from non-relativistic turbulence; and (5) applications of our numerical results to neutron star magnetospheres show that the strong (or moderately weak) turbulent cascade can explain the X-ray radiation of the Vela pulsar. Our study is of great significance for understanding energy transfer, magnetic field evolution, and particle acceleration mechanisms in extreme astrophysical environments.

Star clusters are well known for their dynamical interactions, an outcome of their high stellar densities; in this paper we use multiwavelength observations to search for the unique outcomes of these interactions in three nearby Galactic open clusters: IC 2602 (30 Myr), NGC 2632 (750 Myr) and M67 (4 Gyr). We compared X-ray observations from all-sky surveys like eROSITA, plus archival observations from Chandra X-ray Observatory, survey radio observations from ASKAP's Evolutionary Map of the Universe survey plus archival VLA observations, in conjunction with new cluster catalogs with Gaia. From X-ray, we found 77 X-ray sources likely associated with IC 2602, 31 X-ray sources in NGC 2632, and 31 near M67's central regions. We were further able to classify these X-ray sources based on their optical variability and any radio emission. Three IC 2602 X-ray sources had radio counterparts, which are likely all chromospherically active binary stars. We also identified luminous radio and X-ray variability from a spectroscopic triple system in M67, WOCS 3012/S1077, which is either consistent with a quiescent black hole binary, or due to an active binary stellar system. A recent population study of optical variables by Anderson & Hunt 2025 shows that the population of optical variables in open clusters clearly changes over cluster age; this pilot study gives evidence that the X-ray population also changes with time, and demonstrates the need for a broader multiwavelength study of Galactic open clusters.