Diffusion Language Models (DLMs) incur high inference cost due to iterative denoising, motivating efficient pruning. Existing pruning heuristics largely inherited from autoregressive (AR) LLMs, typically preserve attention sink tokens because AR sinks serve as stable global anchors. We show that this assumption does not hold for DLMs: the attention-sink position exhibits substantially higher variance over the full generation trajectory (measured by how the dominant sink locations shift across timesteps), indicating that sinks are often transient and less structurally essential than in AR models. Based on this observation, we propose ${\bf \texttt{Sink-Aware Pruning}}$, which automatically identifies and prunes unstable sinks in DLMs (prior studies usually keep sinks for AR LLMs). Without retraining, our method achieves a better quality-efficiency trade-off and outperforms strong prior pruning baselines under matched compute. Our code is available at https://github.com/VILA-Lab/Sink-Aware-Pruning.

HIPE-2026 is a CLEF evaluation lab dedicated to person-place relation extraction from noisy, multilingual historical texts. Building on the HIPE-2020 and HIPE-2022 campaigns, it extends the series toward semantic relation extraction by targeting the task of identifying person--place associations in multiple languages and time periods. Systems are asked to classify relations of two types - $at$ ("Has the person ever been at this place?") and $isAt$ ("Is the person located at this place around publication time?") - requiring reasoning over temporal and geographical cues. The lab introduces a three-fold evaluation profile that jointly assesses accuracy, computational efficiency, and domain generalization. By linking relation extraction to large-scale historical data processing, HIPE-2026 aims to support downstream applications in knowledge-graph construction, historical biography reconstruction, and spatial analysis in digital humanities.

Vision-Language-Action models (VLAs) promise to ground language instructions in robot control, yet in practice often fail to faithfully follow language. When presented with instructions that lack strong scene-specific supervision, VLAs suffer from counterfactual failures: they act based on vision shortcuts induced by dataset biases, repeatedly executing well-learned behaviors and selecting objects frequently seen during training regardless of language intent. To systematically study it, we introduce LIBERO-CF, the first counterfactual benchmark for VLAs that evaluates language following capability by assigning alternative instructions under visually plausible LIBERO layouts. Our evaluation reveals that counterfactual failures are prevalent yet underexplored across state-of-the-art VLAs. We propose Counterfactual Action Guidance (CAG), a simple yet effective dual-branch inference scheme that explicitly regularizes language conditioning in VLAs. CAG combines a standard VLA policy with a language-unconditioned Vision-Action (VA) module, enabling counterfactual comparison during action selection. This design reduces reliance on visual shortcuts, improves robustness on under-observed tasks, and requires neither additional demonstrations nor modifications to existing architectures or pretrained models. Extensive experiments demonstrate its plug-and-play integration across diverse VLAs and consistent improvements. For example, on LIBERO-CF, CAG improves $π_{0.5}$ by 9.7% in language following accuracy and 3.6% in task success on under-observed tasks using a training-free strategy, with further gains of 15.5% and 8.5%, respectively, when paired with a VA model. In real-world evaluations, CAG reduces counterfactual failures of 9.4% and improves task success by 17.2% on average.

Language Identification (LID) is an important component of many multilingual natural language processing pipelines, where it facilitates corpus curation, training data analysis, and cross-lingual evaluation of large language models. Despite near-perfect performance on high-resource languages, existing systems remain brittle in low-resource and closely related language settings. We introduce UniLID, a simple and efficient LID method based on the UnigramLM tokenization algorithm, leveraging its probabilistic framing, parameter estimation technique and inference strategy. In short, we learn language-conditional unigram distributions over a shared tokenizer vocabulary but treat segmentation as a language-specific phenomenon. Our formulation is data- and compute-efficient, supports incremental addition of new languages without retraining existing models, and can naturally be integrated into existing language model tokenization pipelines. Empirical evaluations against widely used baselines, including fastText, GlotLID, and CLD3, show that UniLID achieves competitive performance on standard benchmarks, substantially improves sample efficiency in low-resource settings - surpassing 70% accuracy with as few as five labeled samples per language - and delivers large gains on fine-grained dialect identification.

We propose a two-stage "Mine and Refine" contrastive training framework for semantic text embeddings to enhance multi-category e-commerce search retrieval. Large scale e-commerce search demands embeddings that generalize to long tail, noisy queries while adhering to scalable supervision compatible with product and policy constraints. A practical challenge is that relevance is often graded: users accept substitutes or complements beyond exact matches, and production systems benefit from clear separation of similarity scores across these relevance strata for stable hybrid blending and thresholding. To obtain scalable policy consistent supervision, we fine-tune a lightweight LLM on human annotations under a three-level relevance guideline and further reduce residual noise via engagement driven auditing. In Stage 1, we train a multilingual Siamese two-tower retriever with a label aware supervised contrastive objective that shapes a robust global semantic space. In Stage 2, we mine hard samples via ANN and re-annotate them with the policy aligned LLM, and introduce a multi-class extension of circle loss that explicitly sharpens similarity boundaries between relevance levels, to further refine and enrich the embedding space. Robustness is additionally improved through additive spelling augmentation and synthetic query generation. Extensive offline evaluations and production A/B tests show that our framework improves retrieval relevance and delivers statistically significant gains in engagement and business impact.

A recent breakthrough [Hirahara and Nanashima, STOC'2024] established that if $\mathsf{NP} \not \subseteq \mathsf{ioP/poly}$, the existence of zero-knowledge with negligible errors for $\mathsf{NP}$ implies the existence of one-way functions (OWFs). In this work, we obtain a characterization of one-way functions from the worst-case complexity of zero-knowledge {\em in the high-error regime}. We say that a zero-knowledge argument is {\em non-trivial} if the sum of its completeness, soundness and zero-knowledge errors is bounded away from $1$. Our results are as follows, assuming $\mathsf{NP} \not \subseteq \mathsf{ioP/poly}$: 1. {\em Non-trivial} Non-Interactive ZK (NIZK) arguments for $\mathsf{NP}$ imply the existence of OWFs. Using known amplification techniques, this result also provides an unconditional transformation from weak to standard NIZK proofs for all meaningful error parameters. 2. We also generalize to the interactive setting: {\em Non-trivial} constant-round public-coin zero-knowledge arguments for $\mathsf{NP}$ imply the existence of OWFs, and therefore also (standard) four-message zero-knowledge arguments for $\mathsf{NP}$. Prior to this work, one-way functions could be obtained from NIZKs that had constant zero-knowledge error $ε_{zk}$ and soundness error $ε_{s}$ satisfying $ε_{zk} + \sqrt{ε_{s}} < 1$ [Chakraborty, Hulett and Khurana, CRYPTO'2025]. However, the regime where $ε_{zk} + \sqrt{ε_{s}} \geq 1$ remained open. This work closes the gap, and obtains new implications in the interactive setting. Our results and techniques could be useful stepping stones in the quest to construct one-way functions from worst-case hardness.

We initiate a systematic study of pseudo-deterministic quantum algorithms. These are quantum algorithms that, for any input, output a canonical solution with high probability. Focusing on the query complexity model, our main contributions include the following complexity separations, which require new lower bound techniques specifically tailored to pseudo-determinism: - We exhibit a problem, Avoid One Encrypted String (AOES), whose classical randomized query complexity is $O(1)$ but is maximally hard for pseudo-deterministic quantum algorithms ($Ω(N)$ query complexity). - We exhibit a problem, Quantum-Locked Estimation (QL-Estimation), for which pseudo-deterministic quantum algorithms admit an exponential speed-up over classical pseudo-deterministic algorithms ($O(\log(N))$ vs. $Θ(\sqrt{N})$), while the randomized query complexity is $O(1)$. Complementing these separations, we show that for any total problem $R$, pseudo-deterministic quantum algorithms admit at most a quintic advantage over deterministic algorithms, i.e., $D(R) = \tilde O(psQ(R)^5)$. On the algorithmic side, we identify a class of quantum search problems that can be made pseudo-deterministic with small overhead, including Grover search, element distinctness, triangle finding, $k$-sum, and graph collision.

Black-box adversarial attacks on Large Vision-Language Models (LVLMs) are challenging due to missing gradients and complex multimodal boundaries. While prior state-of-the-art transfer-based approaches like M-Attack perform well using local crop-level matching between source and target images, we find this induces high-variance, nearly orthogonal gradients across iterations, violating coherent local alignment and destabilizing optimization. We attribute this to (i) ViT translation sensitivity that yields spike-like gradients and (ii) structural asymmetry between source and target crops. We reformulate local matching as an asymmetric expectation over source transformations and target semantics, and build a gradient-denoising upgrade to M-Attack. On the source side, Multi-Crop Alignment (MCA) averages gradients from multiple independently sampled local views per iteration to reduce variance. On the target side, Auxiliary Target Alignment (ATA) replaces aggressive target augmentation with a small auxiliary set from a semantically correlated distribution, producing a smoother, lower-variance target manifold. We further reinterpret momentum as Patch Momentum, replaying historical crop gradients; combined with a refined patch-size ensemble (PE+), this strengthens transferable directions. Together these modules form M-Attack-V2, a simple, modular enhancement over M-Attack that substantially improves transfer-based black-box attacks on frontier LVLMs: boosting success rates on Claude-4.0 from 8% to 30%, Gemini-2.5-Pro from 83% to 97%, and GPT-5 from 98% to 100%, outperforming prior black-box LVLM attacks. Code and data are publicly available at: https://github.com/vila-lab/M-Attack-V2.

Traditional electronic recycling processes suffer from significant resource loss due to inadequate material separation and identification capabilities, limiting material recovery. We present A.R.I.S. (Automated Recycling Identification System), a low-cost, portable sorter for shredded e-waste that addresses this efficiency gap. The system employs a YOLOx model to classify metals, plastics, and circuit boards in real time, achieving low inference latency with high detection accuracy. Experimental evaluation yielded 90% overall precision, 82.2% mean average precision (mAP), and 84% sortation purity. By integrating deep learning with established sorting methods, A.R.I.S. enhances material recovery efficiency and lowers barriers to advanced recycling adoption. This work complements broader initiatives in extending product life cycles, supporting trade-in and recycling programs, and reducing environmental impact across the supply chain.

Feature engineering remains a critical yet challenging bottleneck in machine learning, particularly for tabular data, as identifying optimal features from an exponentially large feature space traditionally demands substantial domain expertise. To address this challenge, we introduce FAMOSE (Feature AugMentation and Optimal Selection agEnt), a novel framework that leverages the ReAct paradigm to autonomously explore, generate, and refine features while integrating feature selection and evaluation tools within an agent architecture. To our knowledge, FAMOSE represents the first application of an agentic ReAct framework to automated feature engineering, especially for both regression and classification tasks. Extensive experiments demonstrate that FAMOSE is at or near the state-of-the-art on classification tasks (especially tasks with more than 10K instances, where ROC-AUC increases 0.23% on average), and achieves the state-of-the-art for regression tasks by reducing RMSE by 2.0% on average, while remaining more robust to errors than other algorithms. We hypothesize that FAMOSE's strong performance is because ReAct allows the LLM context window to record (via iterative feature discovery and evaluation steps) what features did or did not work. This is similar to a few-shot prompt and guides the LLM to invent better, more innovative features. Our work offers evidence that AI agents are remarkably effective in solving problems that require highly inventive solutions, such as feature engineering.

Retrieving user-specified objects from complex scenes remains a challenging task, especially when queries are ambiguous or involve multiple similar objects. Existing open-vocabulary detectors operate in a one-shot manner, lacking the ability to refine predictions based on user feedback. To address this, we propose IntRec, an interactive object retrieval framework that refines predictions based on user feedback. At its core is an Intent State (IS) that maintains dual memory sets for positive anchors (confirmed cues) and negative constraints (rejected hypotheses). A contrastive alignment function ranks candidate objects by maximizing similarity to positive cues while penalizing rejected ones, enabling fine-grained disambiguation in cluttered scenes. Our interactive framework provides substantial improvements in retrieval accuracy without additional supervision. On LVIS, IntRec achieves 35.4 AP, outperforming OVMR, CoDet, and CAKE by +2.3, +3.7, and +0.5, respectively. On the challenging LVIS-Ambiguous benchmark, it improves performance by +7.9 AP over its one-shot baseline after a single corrective feedback, with less than 30 ms of added latency per interaction.

Let $P$ be a set of $n$ points in the plane, not all on a line, each colored \emph{red} or \emph{blue}. The classical Motzkin--Rabin theorem guarantees the existence of a \emph{monochromatic} line. Motivated by the seminal work of Green and Tao (2013) on the Sylvester-Gallai theorem, we investigate the quantitative and structural properties of monochromatic geometric objects, such as lines, circles, and conics. We first show that if no line contains more than three points, then for all sufficiently large $n$ there are at least $n^{2}/24 - O(1)$ monochromatic lines. We then show a converse of a theorem of Jamison (1986): Given $n\ge 6$ blue points and $n$ red points, if the blue points lie on a conic and every line through two blue points contains a red point, then all red points are collinear. We also settle the smallest nontrivial case of a conjecture of Milićević (2018) by showing that if we have $5$ blue points with no three collinear and $5$ red points, if the blue points lie on a conic and every line through two blue points contains a red point, then all $10$ points lie on a cubic curve. Further, we analyze the random setting and show that, for any non-collinear set of $n\ge 10$ points independently colored red or blue, the expected number of monochromatic lines is minimized by the \emph{near-pencil} configuration. Finally, we examine monochromatic circles and conics, and exhibit several natural families in which no such monochromatic objects exist.

Existing methods for Virtual Try-On (VTON) often struggle to preserve fine garment details, especially in unpaired settings where accurate person-garment correspondence is required. These methods do not explicitly enforce person-garment alignment and fail to explain how correspondence emerges within Diffusion Transformers (DiTs). In this paper, we first analyze full 3D attention in DiT-based architecture and reveal that the person-garment correspondence critically depends on precise person-garment query-key matching within the full 3D attention. Building on this insight, we then introduce CORrespondence ALignment (CORAL), a DiT-based framework that explicitly aligns query-key matching with robust external correspondences. CORAL integrates two complementary components: a correspondence distillation loss that aligns reliable matches with person-garment attention, and an entropy minimization loss that sharpens the attention distribution. We further propose a VLM-based evaluation protocol to better reflect human preference. CORAL consistently improves over the baseline, enhancing both global shape transfer and local detail preservation. Extensive ablations validate our design choices.

Learning time series foundation models has been shown to be a promising approach for zero-shot time series forecasting across diverse time series domains. Insofar as scaling has been a critical driver of performance of foundation models in other modalities such as language and vision, much recent work on time series foundation modeling has focused on scaling. This has resulted in time series foundation models with hundreds of millions of parameters that are, while performant, inefficient and expensive to use in practice. This paper describes a simple recipe for learning efficient foundation models for zero-shot time series forecasting that are orders of magnitude smaller. We show that large-scale transformers are not necessary: small hybrid models that interleave long convolution and linear RNN layers (in particular DeltaNet layers) can match the performance of larger transformer-based models while being more than a hundred times smaller. We also describe several data augmentation and inference strategies that further improve performance. This recipe results in Reverso, a family of efficient time series foundation models for zero-shot forecasting that significantly push the performance-efficiency Pareto frontier.

Reasoning with LLMs increasingly unfolds inside a broader verification loop. Internally, systems use cheap checks, such as self-consistency or proxy rewards, which we call weak verification. Externally, users inspect outputs and steer the model through feedback until results are trustworthy, which we call strong verification. These signals differ sharply in cost and reliability: strong verification can establish trust but is resource-intensive, while weak verification is fast and scalable but noisy and imperfect. We formalize this tension through weak--strong verification policies, which decide when to accept or reject based on weak verification and when to defer to strong verification. We introduce metrics capturing incorrect acceptance, incorrect rejection, and strong-verification frequency. Over population, we show that optimal policies admit a two-threshold structure and that calibration and sharpness govern the value of weak verifiers. Building on this, we develop an online algorithm that provably controls acceptance and rejection errors without assumptions on the query stream, the language model, or the weak verifier.

Modern big-data systems generate massive, heterogeneous, and geographically dispersed streams that are large-scale and privacy-sensitive, making centralization challenging. While federated learning (FL) provides a privacy-enhancing training mechanism, it assumes a static data flow and learns a collaborative model over multiple rounds, making learning with \textit{incremental} data challenging in limited-communication scenarios. This paper presents One-Shot Incremental Federated Learning (OSI-FL), the first FL framework that addresses the dual challenges of communication overhead and catastrophic forgetting. OSI-FL communicates category-specific embeddings, devised by a frozen vision-language model (VLM) from each client in a single communication round, which a pre-trained diffusion model at the server uses to synthesize new data similar to the client's data distribution. The synthesized samples are used on the server for training. However, two challenges still persist: i) tasks arriving incrementally need to retrain the global model, and ii) as future tasks arrive, retraining the model introduces catastrophic forgetting. To this end, we augment training with Selective Sample Retention (SSR), which identifies and retains the top-p most informative samples per category and task pair based on sample loss. SSR bounds forgetting by ensuring that representative retained samples are incorporated into training in further iterations. The experimental results indicate that OSI-FL outperforms baselines, including traditional and one-shot FL approaches, in both class-incremental and domain-incremental scenarios across three benchmark datasets.

While emerging Persian NLP benchmarks have expanded into pragmatics and politeness, they rarely distinguish between memorized cultural facts and the ability to reason about implicit social norms. We introduce DivanBench, a diagnostic benchmark focused on superstitions and customs, arbitrary, context-dependent rules that resist simple logical deduction. Through 315 questions across three task types (factual retrieval, paired scenario verification, and situational reasoning), we evaluate seven Persian LLMs and reveal three critical failures: most models exhibit severe acquiescence bias, correctly identifying appropriate behaviors but failing to reject clear violations; continuous Persian pretraining amplifies this bias rather than improving reasoning, often degrading the model's ability to discern contradictions; and all models show a 21\% performance gap between retrieving factual knowledge and applying it in scenarios. These findings demonstrate that cultural competence requires more than scaling monolingual data, as current models learn to mimic cultural patterns without internalizing the underlying schemas.

LLM-based agents show promise for automating penetration testing, yet reported performance varies widely across systems and benchmarks. We analyze 28 LLM-based penetration testing systems and evaluate five representative implementations across three benchmarks of increasing complexity. Our analysis reveals two distinct failure modes: Type A failures stem from capability gaps (missing tools, inadequate prompts) that engineering readily addresses, while Type B failures persist regardless of tooling due to planning and state management limitations. We show that Type B failures share a root cause that is largely invariant to the underlying LLM: agents lack real-time task difficulty estimation. As a result, agents misallocate effort, over-commit to low-value branches, and exhaust context before completing attack chains. Based on this insight, we present Excalibur, a penetration testing agent that couples strong tooling with difficulty-aware planning. A Tool and Skill Layer eliminates Type A failures through typed interfaces and retrieval-augmented knowledge. A Task Difficulty Assessment (TDA) mechanism addresses Type B failures by estimating tractability through four measurable dimensions (horizon estimation, evidence confidence, context load, and historical success) and uses these estimates to guide exploration-exploitation decisions within an Evidence-Guided Attack Tree Search (EGATS) framework. Excalibur achieves up to 91% task completion on CTF benchmarks with frontier models (39 to 49% relative improvement over baselines) and compromises 4 of 5 hosts on the GOAD Active Directory environment versus 2 by prior systems. These results show that difficulty-aware planning yields consistent end-to-end gains across models and addresses a limitation that model scaling alone does not eliminate.

This paper presents a method to assess the pointing and image motion performance of optical payloads in the presence of image displacement (shift), smear, and jitter. The method assumes the motion is a stationary random process over an image exposure interval. Displacement, smear, and jitter covariances are computed from the solution to a Lyapunov differential equation. These covariances parameterize statistical image motion modulation transfer functions (MTFs), and they can be used to verify pointing and image motion MTF requirements. The method in the present paper extends a previous method to include smear, as well as displacement, and hence jitter. The approach in the present paper also leads, as a special case, to a more efficient method to compute the displacement covariance than the previous method. Numerical examples illustrate the proposed method.

Emerging IoT applications are transitioning from battery-powered to grid-powered nodes. DRP, a contention-based data dissemination protocol, was developed for these applications. Traditional contention-based protocols resolve collisions through control packet exchanges, significantly reducing goodput. DRP mitigates this issue by employing a distributed delay timer mechanism that assigns transmission-start delays based on the average link quality between a sender and its children, prioritizing highly connected nodes for early transmission. However, our in-field experiments reveal that DRP is unable to accommodate real-world link quality fluctuations, leading to overlapping transmissions from multiple senders. This overlap triggers CSMA's random back-off delays, ultimately degrading the goodput performance. To address these shortcomings, we first conduct a theoretical analysis that characterizes the design requirements induced by real-world link quality fluctuations and DRP's passive acknowledgments. Guided by this analysis, we design EDRP, which integrates two novel components: (i) Link-Quality Aware CSMA (LQ-CSMA) and (ii) a Machine Learning-based Block Size Selection (ML-BSS) algorithm for rateless codes. LQ-CSMA dynamically restricts the back-off delay range based on real-time link quality estimates, ensuring that nodes with stronger connectivity experience shorter delays. ML-BSS algorithm predicts future link quality conditions and optimally adjusts the block size for rateless coding, reducing overhead and enhancing goodput. In-field evaluations of EDRP demonstrate an average goodput improvement of 39.43\% than the competing protocols.

Big data scenarios, where massive, heterogeneous datasets are distributed across clients, demand scalable, privacy-preserving learning methods. Federated learning (FL) enables decentralized training of machine learning (ML) models across clients without data centralization. Decentralized training, however, introduces a computational burden on client devices. U-shaped federated split learning (UFSL) offloads a fraction of the client computation to the server while keeping both data and labels on the clients' side. However, the intermediate representations (i.e., smashed data) shared by clients with the server are prone to exposing clients' private data. To reduce exposure of client data through intermediate data representations, this work proposes k-anonymous differentially private UFSL (KD-UFSL), which leverages privacy-enhancing techniques such as microaggregation and differential privacy to minimize data leakage from the smashed data transferred to the server. We first demonstrate that an adversary can access private client data from intermediate representations via a data-reconstruction attack, and then present a privacy-enhancing solution, KD-UFSL, to mitigate this risk. Our experiments indicate that, alongside increasing the mean squared error between the actual and reconstructed images by up to 50% in some cases, KD-UFSL also decreases the structural similarity between them by up to 40% on four benchmarking datasets. More importantly, KD-UFSL improves privacy while preserving the utility of the global model. This highlights its suitability for large-scale big data applications where privacy and utility must be balanced.

Driven by scientific and industry ambition, HPC and AI applications such as operational Numerical Weather Prediction (NWP) require processing and storing ever-increasing data volumes as fast as possible. Whilst POSIX distributed file systems and NVMe SSDs are currently a common HPC storage configuration providing I/O to applications, new storage solutions have proliferated or gained traction over the last decade with potential to address performance limitations POSIX file systems manifest at scale for certain I/O workloads. This work has primarily aimed to assess the suitability and performance of two object storage systems -namely DAOS and Ceph- for the ECMWF's operational NWP as well as for HPC and AI applications in general. New software-level adapters have been developed which enable the ECMWF's NWP to leverage these systems, and extensive I/O benchmarking has been conducted on a few computer systems, comparing the performance delivered by the evaluated object stores to that of equivalent Lustre file system deployments on the same hardware. Challenges of porting to object storage and its benefits with respect to the traditional POSIX I/O approach have been discussed and, where possible, domain-agnostic performance analysis has been conducted, leading to insight also of relevance to I/O practitioners and the broader HPC community. DAOS and Ceph have both demonstrated excellent performance, but DAOS stood out relative to Ceph and Lustre, providing superior scalability and flexibility for applications to perform I/O at scale as desired. This sets a promising outlook for DAOS and object storage, which might see greater adoption at HPC centres in the years to come, although not necessarily implying a shift away from POSIX-like I/O.

Regulatory limits on Maximum Permissible Exposure (MPE) require handheld devices to reduce transmit power when operated near the user's body. Current proximity sensors provide only binary detection, triggering conservative power back-off that degrades link quality. If the device could measure its distance from the body, transmit power could be adjusted proportionally, improving throughput while maintaining compliance. This paper develops a device-centric integrated sensing and communication (ISAC) method for the device to measure this distance. The uplink communication waveform is exploited for sensing, and the natural motion of the user's hand creates a virtual aperture that provides the angular resolution necessary for localization. Virtual aperture processing requires precise knowledge of the device trajectory, which in this scenario is opportunistic and unknown. One can exploit onboard inertial sensors to estimate the device trajectory; however, the inertial sensors accuracy is not sufficient. To address this, we develop an autofocus algorithm based on extended Kalman filtering that jointly tracks the trajectory and compensates residual errors using phase observations from strong scatterers. The Bayesian Cramér-Rao bound for localization is derived under correlated inertial errors. Numerical results at 28GHz demonstrate centimeter-level accuracy with realistic sensor parameters.

The proliferation of Large Language Models (LLMs) necessitates efficient mechanisms to distinguish machine-generated content from human text. While statistical watermarking has emerged as a promising solution, existing methods suffer from two critical limitations: the lack of a principled approach for selecting sampling distributions and the reliance on fixed-horizon hypothesis testing, which precludes valid early stopping. In this paper, we bridge this gap by developing the first e-value-based watermarking framework, Anchored E-Watermarking, that unifies optimal sampling with anytime-valid inference. Unlike traditional approaches where optional stopping invalidates Type-I error guarantees, our framework enables valid, anytime-inference by constructing a test supermartingale for the detection process. By leveraging an anchor distribution to approximate the target model, we characterize the optimal e-value with respect to the worst-case log-growth rate and derive the optimal expected stopping time. Our theoretical claims are substantiated by simulations and evaluations on established benchmarks, showing that our framework can significantly enhance sample efficiency, reducing the average token budget required for detection by 13-15% relative to state-of-the-art baselines.

PDEs are central to scientific and engineering modeling, yet designing accurate numerical solvers typically requires substantial mathematical expertise and manual tuning. Recent neural network-based approaches improve flexibility but often demand high computational cost and suffer from limited interpretability. We introduce \texttt{AutoNumerics}, a multi-agent framework that autonomously designs, implements, debugs, and verifies numerical solvers for general PDEs directly from natural language descriptions. Unlike black-box neural solvers, our framework generates transparent solvers grounded in classical numerical analysis. We introduce a coarse-to-fine execution strategy and a residual-based self-verification mechanism. Experiments on 24 canonical and real-world PDE problems demonstrate that \texttt{AutoNumerics} achieves competitive or superior accuracy compared to existing neural and LLM-based baselines, and correctly selects numerical schemes based on PDE structural properties, suggesting its viability as an accessible paradigm for automated PDE solving.

Molecular generation with diffusion models has emerged as a promising direction for AI-driven drug discovery and materials science. While graph diffusion models have been widely adopted due to the discrete nature of 2D molecular graphs, existing models suffer from low chemical validity and struggle to meet the desired properties compared to 1D modeling. In this work, we introduce MolHIT, a powerful molecular graph generation framework that overcomes long-standing performance limitations in existing methods. MolHIT is based on the Hierarchical Discrete Diffusion Model, which generalizes discrete diffusion to additional categories that encode chemical priors, and decoupled atom encoding that splits the atom types according to their chemical roles. Overall, MolHIT achieves new state-of-the-art performance on the MOSES dataset with near-perfect validity for the first time in graph diffusion, surpassing strong 1D baselines across multiple metrics. We further demonstrate strong performance in downstream tasks, including multi-property guided generation and scaffold extension.

Music generation has advanced markedly through multimodal deep learning, enabling models to synthesize audio from text and, more recently, from images. However, existing image-conditioned systems suffer from two fundamental limitations: (i) they are typically trained on natural photographs, limiting their ability to capture the richer semantic, stylistic, and cultural content of artworks; and (ii) most rely on an image-to-text conversion stage, using language as a semantic shortcut that simplifies conditioning but prevents direct visual-to-audio learning. Motivated by these gaps, we introduce ArtSound, a large-scale multimodal dataset of 105,884 artwork-music pairs enriched with dual-modality captions, obtained by extending ArtGraph and the Free Music Archive. We further propose ArtToMus, the first framework explicitly designed for direct artwork-to-music generation, which maps digitized artworks to music without image-to-text translation or language-based semantic supervision. The framework projects visual embeddings into the conditioning space of a latent diffusion model, enabling music synthesis guided solely by visual information. Experimental results show that ArtToMus generates musically coherent and stylistically consistent outputs that reflect salient visual cues of the source artworks. While absolute alignment scores remain lower than those of text-conditioned systems-as expected given the substantially increased difficulty of removing linguistic supervision-ArtToMus achieves competitive perceptual quality and meaningful cross-modal correspondence. This work establishes direct visual-to-music generation as a distinct and challenging research direction, and provides resources that support applications in multimedia art, cultural heritage, and AI-assisted creative practice. Code and dataset will be publicly released upon acceptance.

We study the topology of the loss landscape of one-hidden-layer ReLU networks under overparameterization. On the theory side, we (i) prove that for convex $L$-Lipschitz losses with an $\ell_1$-regularized second layer, every pair of models at the same loss level can be connected by a continuous path within an arbitrarily small loss increase $ε$ (extending a known result for the quadratic loss); (ii) obtain an asymptotic upper bound on the energy gap $ε$ between local and global minima that vanishes as the width $m$ grows, implying that the landscape flattens and sublevel sets become connected in the limit. Empirically, on a synthetic Moons dataset and on the Wisconsin Breast Cancer dataset, we measure pairwise energy gaps via Dynamic String Sampling (DSS) and find that wider networks exhibit smaller gaps; in particular, a permutation test on the maximum gap yields $p_{perm}=0$, indicating a clear reduction in the barrier height.

Rigorously evaluating machine intelligence against the broad spectrum of human general intelligence has become increasingly important and challenging in this era of rapid technological advance. Conventional AI benchmarks typically assess only narrow capabilities in a limited range of human activity. Most are also static, quickly saturating as developers explicitly or implicitly optimize for them. We propose that a more promising way to evaluate human-like general intelligence in AI systems is through a particularly strong form of general game playing: studying how and how well they play and learn to play \textbf{all conceivable human games}, in comparison to human players with the same level of experience, time, or other resources. We define a "human game" to be a game designed by humans for humans, and argue for the evaluative suitability of this space of all such games people can imagine and enjoy -- the "Multiverse of Human Games". Taking a first step towards this vision, we introduce the AI GameStore, a scalable and open-ended platform that uses LLMs with humans-in-the-loop to synthesize new representative human games, by automatically sourcing and adapting standardized and containerized variants of game environments from popular human digital gaming platforms. As a proof of concept, we generated 100 such games based on the top charts of Apple App Store and Steam, and evaluated seven frontier vision-language models (VLMs) on short episodes of play. The best models achieved less than 10\% of the human average score on the majority of the games, and especially struggled with games that challenge world-model learning, memory and planning. We conclude with a set of next steps for building out the AI GameStore as a practical way to measure and drive progress toward human-like general intelligence in machines.

We present BMC4TimeSec, an end-to-end tool for verifying Timed Security Protocols (TSP) based on SMT-based bounded model checking and multi-agent modelling in the form of Timed Interpreted Systems (TIS) and Timed Interleaved Interpreted Systems (TIIS). In BMC4TimeSec, TSP executions implement the TIS/TIIS environment (join actions, interleaving, delays, lifetimes), and knowledge automata implement the agents (evolution of participant knowledge, including the intruder). The code is publicly available on \href{https://github.com/agazbrzezny/BMC4TimeSec}{GitHub}, as is a \href{https://youtu.be/aNybKz6HwdA}{video} demonstration.