Retrieval augmented generation (RAG) has been widely adopted to help Large Language Models (LLMs) to process tasks involving long documents. However, existing retrieval models are not designed for long document retrieval and fail to address several key challenges of long document retrieval, including context-awareness, causal dependence, and scope of retrieval. In this paper, we proposed AttentionRetriever, a novel long document retrieval model that leverages attention mechanism and entity-based retrieval to build context-aware embeddings for long document and determine the scope of retrieval. With extensive experiments, we found AttentionRetriever is able to outperform existing retrieval models on long document retrieval datasets by a large margin while remaining as efficient as dense retrieval models.

Test-time scaling has become a standard way to improve performance and boost reliability of neural network models. However, its behavior on agentic, multi-step tasks remains less well-understood: small per-step errors can compound over long horizons; and we find that naive policies that uniformly increase sampling show diminishing returns. In this work, we present CATTS, a simple technique for dynamically allocating compute for multi-step agents. We first conduct an empirical study of inference-time scaling for web agents. We find that uniformly increasing per-step compute quickly saturates in long-horizon environments. We then investigate stronger aggregation strategies, including an LLM-based Arbiter that can outperform naive voting, but that can overrule high-consensus decisions. We show that uncertainty statistics derived from the agent's own vote distribution (entropy and top-1/top-2 margin) correlate with downstream success and provide a practical signal for dynamic compute allocation. Based on these findings, we introduce Confidence-Aware Test-Time Scaling (CATTS), which uses vote-derived uncertainty to allocate compute only when decisions are genuinely contentious. CATTS improves performance on WebArena-Lite and GoBrowse by up to 9.1% over React while using up to 2.3x fewer tokens than uniform scaling, providing both efficiency gains and an interpretable decision rule.

We propose an optimization-informed deep neural network approach, named iUzawa-Net, aiming for the first solver that enables real-time solutions for a class of nonsmooth optimal control problems of linear partial differential equations (PDEs). The iUzawa-Net unrolls an inexact Uzawa method for saddle point problems, replacing classical preconditioners and PDE solvers with specifically designed learnable neural networks. We prove universal approximation properties and establish the asymptotic $\varepsilon$-optimality for the iUzawa-Net, and validate its promising numerical efficiency through nonsmooth elliptic and parabolic optimal control problems. Our techniques offer a versatile framework for designing and analyzing various optimization-informed deep learning approaches to optimal control and other PDE-constrained optimization problems. The proposed learning-to-control approach synergizes model-based optimization algorithms and data-driven deep learning techniques, inheriting the merits of both methodologies.

Real-time video generation with Diffusion Transformers is bottlenecked by the quadratic cost of 3D self-attention, especially in real-time regimes that are both few-step and autoregressive, where errors compound across time and each denoising step must carry substantially more information. In this setting, we find that prior sparse-attention approximations break down, despite showing strong results for bidirectional, many-step diffusion. Specifically, we observe that video attention is not reliably sparse, but instead combines pronounced periodic structure driven by spatiotemporal position with dynamic, sparse semantic correspondences and dense mixing, exceeding the representational capacity of even oracle top-k attention. Building on this insight, we propose Monarch-RT, a structured attention parameterization for video diffusion models that factorizes attention using Monarch matrices. Through appropriately aligned block structure and our extended tiled Monarch parameterization, we achieve high expressivity while preserving computational efficiency. We further overcome the overhead of parameterization through finetuning, with custom Triton kernels. We first validate the high efficacy of Monarch-RT over existing sparse baselines designed only for bidirectional models. We further observe that Monarch-RT attains up to 95% attention sparsity with no loss in quality when applied to the state-of-the-art model Self-Forcing, making Monarch-RT a pioneering work on highly-capable sparse attention parameterization for real-time video generation. Our optimized implementation outperforms FlashAttention-2, FlashAttention-3, and FlashAttention-4 kernels on Nvidia RTX 5090, H100, and B200 GPUs respectively, providing kernel speedups in the range of 1.4-11.8X. This enables us, for the first time, to achieve true real-time video generation with Self-Forcing at 16 FPS on a single RTX 5090.

Copyright law focuses on whether a new work is "substantially similar" to an existing one, but generative AI can closely imitate style without copying content, a capability now central to ongoing litigation. We argue that existing definitions of infringement are ill-suited to this setting and propose a new criterion: a generative AI output infringes on an existing work if it could not have been generated without that work in its training corpus. To operationalize this definition, we model generative systems as closure operators mapping a corpus of existing works to an output of new works. AI generated outputs are \emph{permissible} if they do not infringe on any existing work according to our criterion. Our results characterize structural properties of permissible generation and reveal a sharp asymptotic dichotomy: when the process of organic creations is light-tailed, dependence on individual works eventually vanishes, so that regulation imposes no limits on AI generation; with heavy-tailed creations, regulation can be persistently constraining.

We study the optimal transmission and scheduling policy for a transmitter (source) communicating with two gossiping receivers aiming at tracking the source's status over time using the age of information (AoI) metric. Gossiping enables local information exchange in a decentralized manner without relying solely on the transmitter's direct communication, which we assume incurs a transmission cost. On the other hand, gossiping may be communicating stale information, necessitating the transmitter's intervention. With communication links having specific success probabilities, we formulate an average-cost Markov Decision Process (MDP) to jointly minimize the sum AoI and transmission cost for such a system in a time-slotted setting. We employ the Relative Value Iteration (RVI) algorithm to evaluate the optimal policy for the transmitter and then prove several structural properties showing that it has an age-difference threshold structure with minimum age activation in the case where gossiping is relatively more reliable. Specifically, direct transmission is optimal only if the minimum AoI of the receivers is large enough and their age difference is below a certain threshold. Otherwise, the transmitter idles to effectively take advantage of gossiping and reduce direct transmission costs. Numerical evaluations demonstrate the significance of our optimal policy compared to multiple baselines. Our result is a first step towards characterizing optimal freshness and transmission cost trade-offs in gossiping networks.

We study Langevin dynamics with noise projected onto the directions orthogonal to an isometric group action. This mathematical model is introduced to shed new light on the effects of symmetry on stochastic gradient descent for over-parametrized models. Our main result identifies a novel form of implicit regularization: when the initial and target density are both invariant under the group action, Langevin dynamics with projected noise is equivalent in law to Langevin dynamics with isotropic diffusion but with an additional drift term proportional to the negative log volume of the group orbit. We prove this result by constructing a coupling of the two processes via a third process on the group itself, and identify the additional drift as the mean curvature of the orbits.

Unit testing is essential for verifying the functional correctness of code modules (e.g., classes, methods), but manually writing unit tests is often labor-intensive and time-consuming. Unit tests generated by tools that employ traditional approaches, such as search-based software testing (SBST), lack readability, naturalness, and practical usability. LLMs have recently provided promising results and become integral to developers' daily practices. Consequently, software repositories now include a mix of human-written tests, LLM-generated tests, and those from tools employing traditional approaches such as SBST. While LLMs' zero-shot capabilities have been widely studied, their few-shot learning potential for unit test generation remains underexplored. Few-shot prompting enables LLMs to learn from examples in the prompt, and automatically retrieving such examples could enhance test suites. This paper empirically investigates how few-shot prompting with different test artifact sources, comprising human, SBST, or LLM, affects the quality of LLM-generated unit tests as program comprehension artifacts and their contribution to improving existing test suites by evaluating not only correctness and coverage but also readability, cognitive complexity, and maintainability in hybrid human-AI codebases. We conducted experiments on HumanEval and ClassEval datasets using GPT-4o, which is integrated into GitHub Copilot and widely used among developers. We also assessed retrieval-based methods for selecting relevant examples. Our results show that LLMs can generate high-quality tests via few-shot prompting, with human-written examples producing the best coverage and correctness. Additionally, selecting examples based on the combined similarity of problem description and code consistently yields the most effective few-shot prompts.

We consider bidding in repeated Bayesian first-price auctions. Bidding algorithms that achieve optimal regret have been extensively studied, but their strategic robustness to the seller's manipulation remains relatively underexplored. Bidding algorithms based on no-swap-regret algorithms achieve both desirable properties, but are suboptimal in terms of statistical and computational efficiency. In contrast, online gradient ascent is the only algorithm that achieves $O(\sqrt{TK})$ regret and strategic robustness [KSS24], where $T$ denotes the number of auctions and $K$ the number of bids. In this paper, we explore whether simple online linear optimization (OLO) algorithms suffice for bidding algorithms with both desirable properties. Our main result shows that sublinear linearized regret is sufficient for strategic robustness. Specifically, we construct simple black-box reductions that convert any OLO algorithm into a strategically robust no-regret bidding algorithm, in both known and unknown value distribution settings. For the known value distribution case, our reduction yields a bidding algorithm that achieves $O(\sqrt{T \log K})$ regret and strategic robustness (with exponential improvement on the $K$-dependence compared to [KSS24]). For the unknown value distribution case, our reduction gives a bidding algorithm with high-probability $O(\sqrt{T (\log K+\log(T/δ)})$ regret and strategic robustness, while removing the bounded density assumption made in [KSS24].

This paper presents a technical curriculum on language-oriented artificial intelligence (AI) in the language and translation (L&T) industry. The curriculum aims to foster domain-specific technical AI literacy among stakeholders in the fields of translation and specialised communication by exposing them to the conceptual and technical/algorithmic foundations of modern language-oriented AI in an accessible way. The core curriculum focuses on 1) vector embeddings, 2) the technical foundations of neural networks, 3) tokenization and 4) transformer neural networks. It is intended to help users develop computational thinking as well as algorithmic awareness and algorithmic agency, ultimately contributing to their digital resilience in AI-driven work environments. The didactic suitability of the curriculum was tested in an AI-focused MA course at the Institute of Translation and Multilingual Communication at TH Koeln. Results suggest the didactic effectiveness of the curriculum, but participant feedback indicates that it should be embedded into higher-level didactic scaffolding - e.g., in the form of lecturer support - in order to enable optimal learning conditions.

Graph neural networks (GNNs) are designed to use attributed graphs to learn representations. Such representations are beneficial in the unsupervised learning of clusters and community detection. Nonetheless, such inference may reveal sensitive groups, clustered systems, or collective behaviors, raising concerns regarding group-level privacy. Community attribution in social and critical infrastructure networks, for example, can expose coordinated asset groups, operational hierarchies, and system dependencies that could be used for profiling or intelligence gathering. We study a defensive setting in which a data publisher (defender) seeks to conceal a community of interest while making limited, utility-aware changes in the network. Our analysis indicates that community concealment is strongly influenced by two quantifiable factors: connectivity at the community boundary and feature similarity between the protected community and adjacent communities. Informed by these findings, we present a perturbation strategy that rewires a set of selected edges and modifies node features to reduce the distinctiveness leveraged by GNN message passing. The proposed method outperforms DICE in our experiments on synthetic benchmarks and real network graphs under identical perturbation budgets. Overall, it achieves median relative concealment improvements of approximately 20-45% across the evaluated settings. These findings demonstrate a mitigation strategy against GNN-based community learning and highlight group-level privacy risks intrinsic to graph learning.

Joint-Embedding Predictive Architectures (JEPAs) aim to learn representations by predicting target embeddings from context embeddings, inducing a scalar compatibility energy in a latent space. In contrast, Quasimetric Reinforcement Learning (QRL) studies goal-conditioned control through directed distance values (cost-to-go) that support reaching goals under asymmetric dynamics. In this short article, we connect these viewpoints by restricting attention to a principled class of JEPA energy functions : intrinsic (least-action) energies, defined as infima of accumulated local effort over admissible trajectories between two states. Under mild closure and additivity assumptions, any intrinsic energy is a quasimetric. In goal-reaching control, optimal cost-to-go functions admit exactly this intrinsic form ; inversely, JEPAs trained to model intrinsic energies lie in the quasimetric value class targeted by QRL. Moreover, we observe why symmetric finite energies are structurally mismatched with one-way reachability, motivating asymmetric (quasimetric) energies when directionality matters.

Open world language conditioned task planning is crucial for robots operating in large-scale household environments. While many recent works attempt to address this problem using Large Language Models (LLMs) via prompting or training, a key challenge remains scalability. Performance often degrades rapidly with increasing environment size, plan length, instruction ambiguity, and constraint complexity. In this work, we propose Any House Any Task (AHAT), a household task planner optimized for long-horizon planning in large environments given ambiguous human instructions. At its core, AHAT utilizes an LLM trained to map task instructions and textual scene graphs into grounded subgoals defined in the Planning Domain Definition Language (PDDL). These subgoals are subsequently solved to generate feasible and optimal long-horizon plans through explicit symbolic reasoning. To enhance the model's ability to decompose complex and ambiguous intentions, we introduce TGPO, a novel reinforcement learning algorithm that integrates external correction of intermediate reasoning traces into Group Relative Policy Optimization (GRPO). Experiments demonstrate that AHAT achieves significant performance gains over state-of-the-art prompting, planning, and learning methods, particularly in human-style household tasks characterized by brief instructions but requiring complex execution plans.

Multi-robot systems require scalable and federated methods to model complex environments under computational and communication constraints. Gaussian Processes (GPs) offer robust probabilistic modeling, but suffer from cubic computational complexity, limiting their applicability in large-scale deployments. To address this challenge, we introduce the pxpGP, a novel distributed GP framework tailored for both centralized and decentralized large-scale multi-robot networks. Our approach leverages sparse variational inference to generate a local compact pseudo-representation. We introduce a sparse variational optimization scheme that bounds local pseudo-datasets and formulate a global scaled proximal-inexact consensus alternating direction method of multipliers (ADMM) with adaptive parameter updates and warm-start initialization. Experiments on synthetic and real-world datasets demonstrate that pxpGP and its decentralized variant, dec-pxpGP, outperform existing distributed GP methods in hyperparameter estimation and prediction accuracy, particularly in large-scale networks.

Latency-critical speech applications (e.g., live transcription, voice commands, and real-time translation) demand low time-to-first-token (TTFT) and high transcription accuracy, particularly on resource-constrained edge devices. Full-attention Transformer encoders remain a strong accuracy baseline for automatic speech recognition (ASR) because every frame can directly attend to every other frame, which resolves otherwise locally ambiguous acoustics using distant lexical context. However, this global dependency incurs quadratic complexity in sequence length, inducing an inherent "encode-the-whole-utterance" latency profile. For streaming use cases, this causes TTFT to grow linearly with utterance length as the encoder must process the entire prefix before any decoder token can be emitted. To better meet the needs of on-device, streaming ASR use cases we introduce Moonshine v2, an ergodic streaming-encoder ASR model that employs sliding-window self-attention to achieve bounded, low-latency inference while preserving strong local context. Our models achieve state of the art word error rates across standard benchmarks, attaining accuracy on-par with models 6x their size while running significantly faster. These results demonstrate that carefully designed local attention is competitive with the accuracy of full attention at a fraction of the size and latency cost, opening new possibilities for interactive speech interfaces on edge devices.

Data mixing -- determining the ratios of data from different domains -- is a first-order concern for training language models (LMs). While existing mixing methods show promise, they fall short when applied during real-world LM development. We present Olmix, a framework that addresses two such challenges. First, the configuration space for developing a mixing method is not well understood -- design choices across existing methods lack justification or consensus and overlook practical issues like data constraints. We conduct a comprehensive empirical study of this space, identifying which design choices lead to a strong mixing method. Second, in practice, the domain set evolves throughout LM development as datasets are added, removed, partitioned, and revised -- a problem setting largely unaddressed by existing works, which assume fixed domains. We study how to efficiently recompute the mixture after the domain set is updated, leveraging information from past mixtures. We introduce mixture reuse, a mechanism that reuses existing ratios and recomputes ratios only for domains affected by the update. Over a sequence of five domain-set updates mirroring real-world LM development, mixture reuse matches the performance of fully recomputing the mix after each update with 74% less compute and improves over training without mixing by 11.6% on downstream tasks.

We introduce Categorical Flow Maps, a flow-matching method for accelerated few-step generation of categorical data via self-distillation. Building on recent variational formulations of flow matching and the broader trend towards accelerated inference in diffusion and flow-based models, we define a flow map towards the simplex that transports probability mass toward a predicted endpoint, yielding a parametrisation that naturally constrains model predictions. Since our trajectories are continuous rather than discrete, Categorical Flow Maps can be trained with existing distillation techniques, as well as a new objective based on endpoint consistency. This continuous formulation also automatically unlocks test-time inference: we can directly reuse existing guidance and reweighting techniques in the categorical setting to steer sampling toward downstream objectives. Empirically, we achieve state-of-the-art few-step results on images, molecular graphs, and text, with strong performance even in single-step generation.

The Adjusted Winner (AW) method is a fundamental procedure for the fair division of indivisible resources between two agents. However, its reliance on splitting resources can lead to practical complications. To address this limitation, we propose an extension of AW that allows the sale of selected resources under a budget constraint, with the proceeds subsequently redistributed, thereby aiming for allocations that remain as equitable as possible. Alongside developing this extended framework, we provide an axiomatic analysis that examines how equitability and envy-freeness are modified in our setting. We then formally define the resulting combinatorial problems, establish their computational complexity, and design a fully polynomial-time approximation scheme (FPTAS) to mitigate their inherent intractability. Finally, we complement our theoretical results with computer-based simulations.

Diffusion alignment adapts pretrained diffusion models to sample from reward-tilted distributions along the denoising trajectory. This process naturally admits a Sequential Monte Carlo (SMC) interpretation, where the denoising model acts as a proposal and reward guidance induces importance weights. Motivated by this view, we introduce Variance Minimisation Policy Optimisation (VMPO), which formulates diffusion alignment as minimising the variance of log importance weights rather than directly optimising a Kullback-Leibler (KL) based objective. We prove that the variance objective is minimised by the reward-tilted target distribution and that, under on-policy sampling, its gradient coincides with that of standard KL-based alignment. This perspective offers a common lens for understanding diffusion alignment. Under different choices of potential functions and variance minimisation strategies, VMPO recovers various existing methods, while also suggesting new design directions beyond KL.

Determining whether neural models internalize physical laws as world models, rather than exploiting statistical shortcuts, remains challenging, especially under out-of-distribution (OOD) shifts. Standard evaluations often test latent capability via downstream adaptation (e.g., fine-tuning or high-capacity probes), but such interventions can change the representations being measured and thus confound what was learned during self-supervised learning (SSL). We propose a non-invasive evaluation protocol, PhyIP. We test whether physical quantities are linearly decodable from frozen representations, motivated by the linear representation hypothesis. Across fluid dynamics and orbital mechanics, we find that when SSL achieves low error, latent structure becomes linearly accessible. PhyIP recovers internal energy and Newtonian inverse-square scaling on OOD tests (e.g., $ρ> 0.90$). In contrast, adaptation-based evaluations can collapse this structure ($ρ\approx 0.05$). These findings suggest that adaptation-based evaluation can obscure latent structures and that low-capacity probes offer a more accurate evaluation of physical world models.

We study the computational cost of differential privacy in terms of memory efficiency. While the trade-off between accuracy and differential privacy is well-understood, the inherent cost of privacy regarding memory use remains largely unexplored. This paper establishes for the first time an unconditional space lower bound for user-level differential privacy by introducing a novel proof technique based on a multi-player communication game. Central to our approach, this game formally links the hardness of low-memory private algorithms to the necessity of ``contribution capping'' -- tracking and limiting the users who disproportionately impact the dataset. We demonstrate that winning this communication game requires transmitting information proportional to the number of over-active users, which translates directly to memory lower bounds. We apply this framework, as an example, to the fundamental problem of estimating the number of distinct elements in a stream and we prove that any private algorithm requires almost $\widetildeΩ(T^{1/3})$ space to achieve certain error rates in a promise variant of the problem. This resolves an open problem in the literature (by Jain et al. NeurIPS 2023 and Cummings et al. ICML 2025) and establishes the first exponential separation between the space complexity of private algorithms and their non-private $\widetilde{O}(1)$ counterparts for a natural statistical estimation task. Furthermore, we show that this communication-theoretic technique generalizes to broad classes of problems, yielding lower bounds for private medians, quantiles, and max-select.

Hybrid architectures combining state-space models with attention have achieved strong efficiency-quality tradeoffs, yet existing approaches either apply attention uniformly or learn static sparse patterns. This misses a key opportunity: \emph{attention demand should decrease over time as recurring patterns become familiar}. We present a surprising finding from analyzing GPT-2 models: \textbf{88\%} of attention operations retrieve information already predictable from the model's hidden state, and this redundancy does \emph{not} decrease during training. Motivated by this observation, we introduce \textbf{\ours{}} (\textbf{C}onsolidation-based \textbf{R}outing for \textbf{A}daptive \textbf{M}emory), a biologically inspired memory consolidation mechanism that gradually distills episodic retrievals into parametric semantic memory. Unlike prior sparse attention methods, \ours{} exhibits \emph{decreasing attention utilization} over training, achieving a \textbf{37.8$\times$} reduction through a sharp phase transition at approximately 3K steps. We prove that this capability is \emph{impossible} without consolidation: any static routing scheme requires $Ω(f \cdot n)$ attention for tasks with recurring patterns of frequency $f$. On our proposed SRCD benchmark, \ours{} achieves \textbf{100\% retrieval accuracy} at 1.6\% attention compute (vs.\ 68\% for baselines), and consolidated patterns transfer to unseen tasks with \textbf{48--52\%} attention reduction without retraining. Remarkably, the learned consolidation dynamics quantitatively match human episodic-to-semantic memory transition curves from cognitive psychology ($γ= 0.43$ vs.\ $γ_{\text{human}} \approx 0.4$--$0.5$). Code and benchmarks are available at [anonymized].

Enterprise documents, such as forms and reports, embed critical information for downstream applications like data archiving, automated workflows, and analytics. Although generalist Vision Language Models (VLMs) perform well on established document understanding benchmarks, their ability to conduct holistic, fine-grained structured extraction across diverse document types and flexible schemas is not well studied. Existing Key Entity Extraction (KEE), Relation Extraction (RE), and Visual Question Answering (VQA) datasets are limited by narrow entity ontologies, simple queries, or homogeneous document types, often overlooking the need for adaptable and structured extraction. To address these gaps, we introduce ExStrucTiny, a new benchmark dataset for structured Information Extraction (IE) from document images, unifying aspects of KEE, RE, and VQA. Built through a novel pipeline combining manual and synthetic human-validated samples, ExStrucTiny covers more varied document types and extraction scenarios. We analyze open and closed VLMs on this benchmark, highlighting challenges such as schema adaptation, query under-specification, and answer localization. We hope our work provides a bedrock for improving generalist models for structured IE in documents.

We propose a geometric model for optimal shape-change-induced motions of slender locomotors, e.g., snakes slithering on sand. In these scenarios, the motion of a body in world coordinates is completely determined by the sequence of shapes it assumes. Specifically, we formulate Lagrangian least-dissipation principles as boundary value problems whose solutions are given by sub-Riemannian geodesics. Notably, our geometric model accounts not only for the energy dissipated by the body's displacement through the environment, but also for the energy dissipated by the animal's metabolism or a robot's actuators to induce shape changes such as bending and stretching, thus capturing overall locomotion efficiency. Our continuous model, together with a consistent time and space discretization, enables numerical computation of sub-Riemannian geodesics for three different types of boundary conditions, i.e., fixing initial and target body, restricting to cyclic motion, or solely prescribing body displacement and orientation. The resulting optimal deformation gaits qualitatively match observed motion trajectories of organisms such as snakes and spermatozoa, as well as known optimality results for low-dimensional systems such as Purcell's swimmers. Moreover, being geometrically less rigid than previous frameworks, our model enables new insights into locomotion mechanisms of, e.g., generalized Purcell's swimmers. The code is publicly available.

AI models have achieved state-of-the-art results in textual reasoning; however, their ability to reason over spatial and relational structures remains a critical bottleneck -- particularly in early-grade maths, which relies heavily on visuals. This paper introduces the visual reasoning benchmark (VRB), a novel dataset designed to evaluate Multimodal Large Language Models (MLLMs) on their ability to solve authentic visual problems from classrooms. This benchmark is built on a set of 701 questions sourced from primary school examinations in Zambia and India, which cover a range of tasks such as reasoning by analogy, pattern completion, and spatial matching. We outline the methodology and development of the benchmark which intentionally uses unedited, minimal-text images to test if models can meet realistic needs of primary education. Our findings reveal a ``jagged frontier'' of capability where models demonstrate better proficiency in static skills such as counting and scaling, but reach a distinct ``spatial ceiling'' when faced with dynamic operations like folding, reflection, and rotation. These weaknesses pose a risk for classroom use on visual reasoning problems, with the potential for incorrect marking, false scaffolding, and reinforcing student misconceptions. Consequently, education-focused benchmarks like the VRB are essential for determining the functional boundaries of multimodal tools used in classrooms.

Biomedical signal classification presents unique challenges due to long sequences, complex temporal dynamics, and multi-scale frequency patterns that are poorly captured by standard transformer architectures. We propose WaveFormer, a transformer architecture that integrates wavelet decomposition at two critical stages: embedding construction, where multi-channel Discrete Wavelet Transform (DWT) extracts frequency features to create tokens containing both time-domain and frequency-domain information, and positional encoding, where Dynamic Wavelet Positional Encoding (DyWPE) adapts position embeddings to signal-specific temporal structure through mono-channel DWT analysis. We evaluate WaveFormer on eight diverse datasets spanning human activity recognition and brain signal analysis, with sequence lengths ranging from 50 to 3000 timesteps and channel counts from 1 to 144. Experimental results demonstrate that WaveFormer achieves competitive performance through comprehensive frequency-aware processing. Our approach provides a principled framework for incorporating frequency-domain knowledge into transformer-based time series classification.

Search-Augmented Generative Engines (SAGE) have emerged as a new paradigm for information access, bridging web-scale retrieval with generative capabilities to deliver synthesized answers. This shift has fundamentally reshaped how web content gains exposure online, giving rise to Search-Augmented Generative Engine Optimization (SAGEO), the practice of optimizing web documents to improve their visibility in AI-generated responses. Despite growing interest, no evaluation environment currently supports comprehensive investigation of SAGEO. Specifically, existing benchmarks lack end-to-end visibility evaluation of optimization strategies, operating on pre-determined candidate documents that abstract away retrieval and reranking preceding generation. Moreover, existing benchmarks discard structural information (e.g., schema markup) present in real web documents, overlooking the rich signals that search systems actively leverage in practice. Motivated by these gaps, we introduce SAGEO Arena, a realistic and reproducible environment for stage-level SAGEO analysis. Our objective is to jointly target search-oriented optimization (SEO) and generation-centric optimization (GEO). To achieve this, we integrate a full generative search pipeline over a large-scale corpus of web documents with rich structural information. Our findings reveal that existing approaches remain largely impractical under realistic conditions and often degrade performance in retrieval and reranking. We also find that structural information helps mitigate these limitations, and that effective SAGEO requires tailoring optimization to each pipeline stage. Overall, our benchmark paves the way for realistic SAGEO evaluation and optimization beyond simplified settings.

The rapid evolution of cyberattacks continues to drive the emergence of unknown (zero-day) threats, posing significant challenges for network intrusion detection systems in Internet of Things (IoT) networks. Existing machine learning and deep learning approaches typically rely on large labeled datasets, payload inspection, or closed-set classification, limiting their effectiveness under data scarcity, encrypted traffic, and distribution shifts. Consequently, detecting unknown attacks in realistic IoT deployments remains difficult. To address these limitations, we propose SiamXBERT, a robust and data-efficient Siamese meta-learning framework empowered by a transformer-based language model for unknown attack detection. The proposed approach constructs a dual-modality feature representation by integrating flow-level and packet-level information, enabling richer behavioral modeling while remaining compatible with encrypted traffic. Through meta-learning, the model rapidly adapts to new attack types using only a small number of labeled samples and generalizes to previously unseen behaviors. Extensive experiments on representative IoT intrusion datasets demonstrate that SiamXBERT consistently outperforms state-of-the-art baselines under both within-dataset and cross-dataset settings while requiring significantly less training data, achieving up to \num{78.8}\% improvement in unknown F1-score. These results highlight the practicality of SiamXBERT for robust unknown attack detection in real-world IoT environments.

We extend the recent analysis of the rate-reliability tradeoff in deterministic identification (DI) to general linear Gaussian channels, marking the first such analysis for channels with continuous output. Because DI provides a framework that can substantially enhance communication efficiency, and since the linear Gaussian model underlies a broad range of physical communication systems, our results offer both theoretical insights and practical relevance for the performance evaluation of DI in future networks. Moreover, the structural parallels observed between the Gaussian and discrete-output cases suggest that similar rate-reliability behaviour may extend to wider classes of continuous channels.

Convex Markov Games (cMGs) were recently introduced as a broad class of multi-agent learning problems that generalize Markov games to settings where strategic agents optimize general utilities beyond additive rewards. While cMGs expand the modeling frontier, their theoretical foundations, particularly the structure of Nash equilibria (NE) and guarantees for learning algorithms, are not yet well understood. In this work, we address these gaps for an extension of cMGs, which we term General Utility Markov Games (GUMGs), capturing new applications requiring coupling between agents' occupancy measures. We prove that in GUMGs, Nash equilibria coincide with the fixed points of projected pseudo-gradient dynamics (i.e., first-order stationary points), enabled by a novel agent-wise gradient domination property. This insight also yields a simple proof of NE existence using Brouwer's fixed-point theorem. We further show the existence of Markov perfect equilibria. Building on this characterization, we establish a policy gradient theorem for GUMGs and design a model-free policy gradient algorithm. For potential GUMGs, we establish iteration complexity guarantees for computing approximate-NE under exact gradients and provide sample complexity bounds in both the generative model and on-policy settings. Our results extend beyond prior work restricted to zero-sum cMGs, providing the first theoretical analysis of common-interest cMGs.