arXiv:2503.14862v3 Announce Type: replace Abstract: Open-vocabulary detectors are proposed to locate and recognize objects in novel classes. However, variations in vision-aware language vocabulary data used for open-vocabulary learning can lead to unfair and unreliable evaluations. Recent evaluation methods have attempted to address this issue by incorporating object properties or adding locations and characteristics to the captions. Nevertheless, since these properties and locations depend on the specific details of the images instead of classes, detectors can not make accurate predictions without precise descriptions provided through human annotation. This paper introduces 3F-OVD, a novel task that extends supervised fine-grained object detection to the open-vocabulary setting. Our task is intuitive and challenging, requiring a deep understanding of Fine-grained captions and careful attention to Fine-grained details in images in order to accurately detect Fine-grained objects. Additionally, due to the scarcity of qualified fine-grained object detection datasets, we have created a new dataset, NEU-171K, tailored for both supervised and open-vocabulary settings. We benchmark state-of-the-art object detectors on our dataset for both settings. Furthermore, we propose a simple yet effective post-processing technique. Our data, annotations and codes are available at https://github.com/tengerye/3FOVD.
Science Journals
arXiv:2604.11284v5 Announce Type: replace Abstract: We present THEIA, a 2.75M-parameter modular neural architecture that learns the complete Kleene three-valued logic (K3) truth table from task data without external symbolic inference or hand-encoded K3 gate primitives. Across 5 seeds it passes all 39 K3 rules at >99% per-rule accuracy. K3 learnability is not the central finding: Transformer baselines also pass all 39 rules, and flat MLPs match THEIA on Phase-1 accuracy within 0.04pp. The contributions are two properties of the learned system. (1) Uncertainty-verdict asymmetric propagation. THEIA preserves Has-Unknown at every upstream boundary (80.0/91.1/90.8/99.7% across Arith/Order/Set/Logic vs. ~52% majority) while final-verdict decodability stays at or below a 73.4% U-vs-non-U oracle reference under linear and nonlinear probes. Activation patching on non-absorbent T->U cases flips 4,898/4,898 OR and 4,719/4,719 AND pairs across 5 seeds, ruling out residual shortcuts. (2) Reliability spectrum under discretized end-to-end training, on tasks decomposable along the engine boundaries. A mod-3 sequential composition task generalizes from 5- to 500-step evaluation at 99.96+-0.04% (5 seeds). Under identical Gumbel-softmax training, flat MLPs collapse to chance by 50 steps; a 2x2 ResMLP grid reaches >=99% on only 3/20 (config, seed) trials; a pre-LN Transformer reaches 99.24+-0.34%. Straight-through discretization prevents 0.999^500 compounding; the architectural separator is sustaining Phase-1 accuracy under Phase-3 training, where flat MLPs fail. Auxiliary: under per-architecture development defaults (not optimizer-controlled), THEIA reaches 12/12 Kleene coverage 6.5x faster than a parameter-comparable 8L Transformer; this narrows to ~3.6x with Transformer-standard tuning and 4.93x with the same recipe on both. Ratios are config-specific, not asymptotic.
arXiv:2505.15370v5 Announce Type: replace Abstract: Understanding information diffusion on X (formerly Twitter) requires accurate modelling of reposting behaviour. Most existing work predicts reposting under in-distribution settings, where training and test data cover the same topics. This paper addresses a more realistic and challenging scenario: out-of-distribution prediction, i.e., forecasting reposting behaviour for new, previously unseen topics. We formulate the task at the individual level - predicting whether a specific user will repost a given post - and systematically compare the predictive power of post-related features, user-related features, and their combination across four representative models: Decision Tree, Multi-Layer Perceptron, BERT, and Qwen. Our experiments show that while post-related features perform well in-distribution, their performance declines drastically for unseen topics, with F1 scores falling to approximately 0.12. In contrast, user-related features - including user profiles, social relations, and historical behaviour - deliver strong and transferable performance, raising the F1 score to over 0.70. These results demonstrate that reposting decisions are largely content-agnostic: they are driven more by stable user characteristics than by the specific content of a post. Our findings highlight the value of user modelling for building robust prediction systems and provide new insights into the mechanisms that enable information to spread across different topics.
arXiv:2505.17338v3 Announce Type: replace Abstract: Photorealistic volumetric rendering of CT scans greatly benefits clinical workflows, yet neural approaches such as Neural Radiance Fields (NeRF) and 3D Gaussian Splatting (3DGS) require prohibitive per-scan optimization (hours for NeRF, about 30 minutes for 3DGS), making them impractical in clinical settings. We propose Render-FM, a feedforward model that eliminates this bottleneck by directly regressing 6D Gaussian Splatting (6DGS) parameters from a CT volume in a single 2.8-second forward pass, a 500x speedup over per-scan optimization. To bridge the domain gap between natural scene reconstruction and medical volumetric rendering, we introduce Anatomy-Guided Priming (AGP), which incorporates segmentation masks and transfer functions as structural and appearance priors, information that existing Gaussian splatting methods overlook. Built on an nnU-Net-inspired 3D U-Net trained on diverse CT scans, Render-FM predicts per-voxel 6DGS parameters and supports immediate real-time rendering. Unlike per-scan methods, it generalizes to unseen anatomies, novel transfer functions, and enables compositional organ visualization with zero additional preparation time. Optional 89-second fine-tuning further improves quality, surpassing per-scan optimized baselines. Project page: https://gaozhongpai.github.io/renderfm/.
arXiv:2510.25444v3 Announce Type: replace Abstract: Convectively Coupled Equatorial Waves (CCEWs) dominate atmospheric variability on timescales of 2--30 days in the Tropics, bringing episodes of widespread heavy precipitation. This study compares the representation of CCEWs and their connection to upscale energy transfer in two Met Office Unified Model simulations of the full tropical channel with identical km-scale resolution for the DYAMOND Summer period. The principal difference between the simulations is that one parametrizes convection (GAL9), while the other (RAL3) is convection permitting. The GAL9 convection scheme acts to remove vertical instability without explicitly representing the resolved-scale circulation associated with convective plumes. We present the first quantitative diagnosis of interscale energy transfer and its relation to CCEWs. This diagnosis is important because upscale energy transfer between convection and large-scale waves may influence accurate simulation of tropical weather systems. The average upper-tropospheric upscale transfer simulated by RAL3 is approximately 50% higher than GAL9. CCEWs are more coherent in RAL3, with an average phase-speed variability 80% higher than observations, compared with 166% higher in GAL. RAL3 also simulates greater upscale energy transfer within waves than GAL9 with a stronger correlation between the interscale energy transfer and equatorial wave winds. Kelvin and Rossby waves are associated with upscale energy transfer from scales 4-8 times smaller than their wavelength, related to active deep convection within a particular sector of the wave phase. Our findings show that the explicit representation of convection has a significant impact on the simulation of upscale energy transfer, and is very likely to be a significant factor in the faithful simulation of convective coupling within CCEWs.
arXiv:2510.24295v2 Announce Type: replace Abstract: As many benchmarks have become saturated, it has become increasingly important to create new datasets that evaluate the generalization capacity of current state-of-the-art models in reasoning. However, designing high-quality reasoning datasets is challenging, as their manual construction is costly, and their automatic generation is unreliable, often leading to synthetic data with limited scope. In this paper, we propose the Minimal Expression-Replacement GEneralization (MERGE) test that evaluates the robustness of reasoning models against non-adversarial variants of existing evaluation datasets. We automatically obtain high-quality variants from the original instances with Minimal Expression REplacement (MERE) generation, which uses Masked Language Models (MLMs) and safeguarding filters. We apply the MERGE test to Natural Language Inference (NLI), a popular task of reasoning. We generate new NLI datasets from two widely used existing ones with the MERE generation and use them to evaluate multiple strong NLI models. The results indicate that both LLMs and fine-tuned NLI models generalize poorly: they struggle to consistently and correctly classify variants minimally different in form and reasoning from the original ones. Further, we also analyze how certain aspects in variant generation, such as the word class and the source MLMs, affect model performance.
arXiv:2602.09303v2 Announce Type: replace Abstract: We propose a physics-informed consistency modeling framework for solving partial differential equations (PDEs) via fast, few-step generative inference. We identify a key stability challenge in physics-constrained consistency training, where PDE residuals can drive the model toward trivial or degenerate solutions, degrading the learned data distribution. To address this, we introduce a structure-preserving two-stage training strategy that decouples distribution learning from physics enforcement by freezing the coefficient decoder during physics-informed fine-tuning. We further propose a two-step residual objective that enforces physical consistency on refined, structurally valid generative trajectories rather than noisy single-step predictions. The resulting framework enables stable, high-fidelity inference for both unconditional generation and forward problems. We demonstrate that forward solutions can be obtained via a projection-based zero-shot inpainting procedure, achieving consistent accuracy of diffusion baselines with orders of magnitude reduction in computational cost.
arXiv:2512.15424v2 Announce Type: replace Abstract: A systematic numerical investigation of flow-regime transitions in the two-dimensional incompressible Navier-Stokes flow past a confined circular cylinder is presented. For a fixed benchmark geometry, we observe a clear empirical correspondence between qualitative changes in steady traction profiles, understood here as the pointwise force density given by the Cauchy stress tensor on the obstacle boundary, and bifurcations in the long-time behavior of the unsteady Navier-Stokes equations. The observed transitions include onset of time-periodic oscillations, the appearance of multiple steady solutions and loss of effective symmetry. The well-known planar Sch\"afer-Turek benchmark is considered for Reynolds numbers up to 500. Several numerical techniques are employed to compute steady solutions, boundary traction profiles, and linear stability spectra such as duality-based approach for traction evaluation, deflation methods for detecting multiple steady states, and both two- and three- dimensional linear stability analyzes. The results suggest that steady boundary traction profiles can serve as a sensitive diagnostic indicator of critical Reynolds numbers at which qualitative changes in flow dynamics occur. This suggests a computationally inexpensive, complementary approach for detecting flow-regime transitions within this benchmark configuration.
arXiv:2512.19842v2 Announce Type: replace Abstract: The complexity and scale of Internet attacks call for distributed, cooperative observatories capable of monitoring malicious traffic across diverse networks. Holoscope is an open, lightweight, and cloud-native platform designed to simplify the deployment and management of telescope (passive) and honeypot (active) sensors. Built upon K3s and WireGuard, Holoscope offers secure connectivity, automated sensor onboarding, and resilient operation even in resource-constrained environments. Through modular design and Infrastructure-as-Code principles, it supports dynamic sensor orchestration, automated recovery, and data processing. We build, deploy, and operate Holoscope across multiple institutions and cloud networks in Europe and Brazil, enabling unified visibility into large-scale attack phenomena while maintaining ease of integration and security compliance.
arXiv:2601.06351v2 Announce Type: replace Abstract: Anticlustering is an NP-hard combinatorial optimization problem that consists of partitioning a set of objects into equal-sized groups called anticlusters such that the objects in the same anticluster are as dissimilar as possible and thereby representative of the entire set of objects. Here we study the case where the dissimilarity metric is the squared Euclidean distance between the respective feature vectors. Applications of Euclidean anticlustering include social studies, cross-validation, creating mini-batches for stochastic gradient descent, and finding balanced K-cut partitions. In particular, machine-learning applications such as mini-batch generation involve million-scale datasets and very large values of K, making scalable anticlustering algorithms essential. We propose a new algorithm, the Assignment-Based Anticlustering (ABA) algorithm, that scales to instances with millions of objects and hundreds of thousands of anticlusters within seconds to minutes, which is far beyond what existing anticlustering methods can manage. We demonstrate here, via an extensive computational study, that our algorithm outperforms existing anticlustering methods in both solution quality and running time. This is so also for anticlustering with categories. For the related problem of balanced K-cut partitioning, our algorithm is superior to the well-known METIS method. The code of our algorithm is available on GitHub.
arXiv:2602.02899v2 Announce Type: replace Abstract: Decentralized training is often regarded as inferior to centralized training because the consensus errors between workers are thought to undermine convergence and generalization. This work challenges this view by introducing decentralized SGD with Adaptive Consensus (DSGD-AC), which uses a time-dependent scaling mechanism to maintain consensus errors throughout the training. We show that adaptive consensus changes the stationary variance of disagreement modes by balancing two effects: it preserves consensus-error magnitude through weaker graph damping while still allowing curvature-dependent damping to shape the disagreement directions. This balance can produce a stronger Hessian-weighted loss-envelope penalty around the deployed model, even when normalized Hessian alignment is weaker than in standard DSGD. Empirical results on image classification show that DSGD-AC reaches flatter solutions and higher test accuracy than standard DSGD and even centralized SGD. Together, these results support consensus errors as a useful implicit regularizer and open a new perspective on the design of decentralized learning algorithms.
arXiv:2603.14109v3 Announce Type: replace Abstract: Millimeter-wave radar enables robust perception in visually degraded environments, yet radar-inertial estimation remains prone to drift: sparse body-frame velocity measurements weakly constrain absolute orientation, leaving IMU biases poorly observable over the short horizons of sliding-window estimators. We propose a tightly coupled, hierarchical radar-inertial factor graph that decouples estimation into a high-rate resetting graph and a persistent global graph. The resetting graph fuses IMU preintegration, radar velocities, and adaptive ZUPT to produce smooth, low-latency odometry for real-time control. The persistent graph maintains a full state (poses, velocities, and biases) via keyframe-based geometric mapping and loop closures. Fully observable biases and their exact covariances are continuously injected from the persistent graph as priors into the resetting graph, anchoring the high-rate estimator against integration drift. Extensive evaluations demonstrate high accuracy and drift-reduced estimation at faster than real-time speeds. Code and datasets will be released upon paper acceptance.
arXiv:2602.14371v2 Announce Type: replace Abstract: The classical definitions extract degrees of freedom (DOF) via $C(\snr)/\log\snr$ and diversity (DIV) via $-\log P_e(\snr)/\log\snr$, using $\log\snr$ as the common gauge for both. These ratios hide a two-step process: first, identify the gauge on which capacity or reliability actually grows; second, normalize the coefficient on that gauge by the appropriate atom. For coherent multiple-input multiple-output (MIMO) both gauges happen to be $\log\snr$ and both atom coefficients happen to be one. This paper shows that the two-step process is necessary outside this calibration case and makes it explicit using a Bhattacharyya-frontier construction. A capacity--packing sandwich theorem shows that fixed-resolution output-law packing and covering recover the capacity gauge, and a binary-endpoint theorem shows that the two-message Bhattacharyya frontier identifies the zero-rate diversity gauge. Endpoint DOF and endpoint DIV are obtained by dividing the raw coefficient on the identified gauge by the corresponding atom coefficient. For fixed deterministic channel matrix~$H$, the capacity gauge is $\log\snr$ with endpoint DOF $T\,\mathrm{rank}(H)$, while the zero-rate diversity gauge is $\snr$ with endpoint DIV $T\sigma_1^2(H)$, making fixed-$H$ a cross-gauge channel. For noncoherent scalar fast fading with $N$~receive antennas, the capacity gauge is $\log\log\snr$ with DOF~$1$, while the zero-rate diversity gauge is $\log\snr$ with endpoint DIV~$N$; the exact load-$r$ frontier gauge is $(\log\snr)^{1-r}$. The framework recovers the same-gauge cases, coherent Rayleigh MIMO and noncoherent block fading, with zero-rate DIV~$MN$. Audit tables separate exact results from lower bounds and open problems.
arXiv:2602.18934v2 Announce Type: replace Abstract: Membership inference attacks (MIAs) threaten the privacy of machine learning models by revealing whether a specific data point was used during training. Existing MIAs often rely on impractical assumptions, such as access to public datasets, shadow models, confidence scores, or knowledge of the training data distribution, making them vulnerable to defenses like confidence masking and adversarial regularization. Label-only MIAs, even under strict constraints, suffer from high query requirements per sample. We propose a cost-effective label-only MIA framework based on transferability and model extraction. By querying the target model $M$ using active sampling, perturbation-based selection, and synthetic data, we extract a functionally similar surrogate model $S$ on which membership inference is performed. This shifts the query overhead to a one-time extraction phase, eliminating repeated queries to $M$. Our method matches the performance of state-of-the-art label-only MIAs while significantly reducing query costs and operating under strict black-box constraints. On benchmark tabular datasets, we show that a query budget equivalent to testing the membership of approximately $1%$ of the training samples is sufficient to extract $S$ and achieve membership inference accuracy within $\pm 1%$ of that obtained when attacking $M$ directly. We also evaluate the effectiveness of standard defenses, including DP-SGD and regularization, proposed for label-only MIAs against our attack. Finally, we present preliminary results extending our framework to deep neural networks trained on image datasets, demonstrating promising transferability and membership inference performance under label-only access while highlighting directions for further optimization.
arXiv:2603.20450v2 Announce Type: replace Abstract: A number of scientific conferences and journals have recently enacted policies that prohibit LLM usage by peer reviewers, except for polishing, paraphrasing, and grammar correction of otherwise human-written reviews. But, are these policies enforceable? To answer this question, we assemble a dataset of peer reviews simulating multiple levels of human-AI collaboration, and evaluate five state-of-the-art detectors, including two commercial systems. Our analysis shows that all detectors misclassify a non-trivial fraction of LLM-polished reviews as AI-generated, thereby risking false accusations of academic misconduct. We further investigate whether peer-review-specific signals, including access to the paper manuscript and the constrained domain of scientific writing, can be leveraged to improve detection. While incorporating such signals yields measurable gains in some settings, we identify limitations in each approach and find that none meets the accuracy standards required for identifying AI use in peer reviews. Importantly, our results suggest that recent public estimates of AI use in peer reviews through the use of AI-text detectors should be interpreted with caution, as current detectors misclassify mixed reviews (collaborative human-AI outputs) as fully AI generated, potentially overstating the extent of policy violations.
arXiv:2603.20510v2 Announce Type: replace Abstract: Language models often lack grounded reasoning capabilities in specialized domains where training data is scarce but bespoke systems excel. We introduce a general framework for distilling expert system reasoning into natural language chain-of-thought explanations, enabling compact models to acquire domain expertise and the ability to generate faithful, grounded explanations. Rather than distilling only final outputs, we capture the full reasoning process, transforming opaque expert computations into transparent, step-by-step explanations. We demonstrate this approach in chess, a canonical reasoning domain where language models continue to underperform. Our 4B parameter model, C1, advances from a near-zero baseline to 48.1\% accuracy, outperforming all open-source models and most frontier proprietary systems. Notably, C1 surpasses its distillation teacher and generates solutions in two orders of magnitude fewer tokens than baselines. Unlike prior neural chess approaches that predict only best moves, C1 generates explainable solutions revealing strategic reasoning. Our pipeline combines supervised fine-tuning and reinforcement learning with theme-balanced data sampling for comprehensive tactical coverage. Master Distillation demonstrates how to inject expert-level knowledge into compact models for under-optimized domains, offering a recipe for unlocking RLVR where LLMs lack sufficient base capabilities.
arXiv:2606.16797v3 Announce Type: replace Abstract: In July 2025, Study Times, sponsored by the Party School of the Central Committee of the CPC, pointed out that 95% of industrial software for R&D and design in China relies on imports, and that 90% of the high-end CAD/CAE/CAM software market is monopolized by European and American giants. This is a typical strategic bottleneck problem. Unlike the visually oriented goal of "visual plausibility" pursued by related sister disciplines such as CV and CG, CAD places greater emphasis on "industrial usability". In CAD, data representation architecture is more foundational than the optimization of network algorithms. This paper first starts from data representation in AI+CAD and reports a classification paradigm and research progress in AI+CAD. Then, using the open-source DeepCAD data representation as an example, it analyzes the pain points of representative AI+CAD work and the gap between such work and real industrial-level parametric feature modeling. Next, by comparison with the open-source WHUCAD data representation, it discusses how its three-level architecture provides fundamental support for industrial-grade parametric feature modeling. Finally, in view of the rapid iteration of the AI wave, large models, and agents, this paper offers an outlook on AI+industrial-grade CAD.
arXiv:2606.20189v3 Announce Type: replace Abstract: Leveraging Vision Foundation Models (VFMs) for camera-to-LiDAR knowledge distillation offers a promising solution to the scarcity of annotated data needed to represent the immense geometric and kinematic diversity of real-world autonomous driving (AD). However, current approaches typically treat VFMs as black-box teachers, relying exclusively on frame-wise feature similarity. Consequently, they do not fully exploit the teacher's layer-wise semantic structure and global context, as well as the rich spatiotemporal information inherent in LiDAR sequences. We propose HilDA, a self-supervised pretraining framework for LiDAR backbones that better captures the semantic what and geometric where needed for driving tasks. HilDA combines hierarchical distillation comprising multi-layer distillation for progressive semantic alignment and global context distillation for scene-level semantics, with a temporal occupancy diffusion objective promoting spatiotemporal consistency. Models pre-trained with HilDA achieve state-of-the-art results on cross-modal distillation benchmarks and outperform models trained via prior distillation approaches on 3D object detection, scene flow, and semantic occupancy prediction. Code available at: https://maxiuw.github.io/hilda.
arXiv:2606.22974v2 Announce Type: replace Abstract: Recent work on preference elicitation in large language models (LLMs) has demonstrated that, when given a series of choices between two outcomes, LLMs reveal a coherent, model-specific utility structure. Notably, this structure often includes preferences that the models' trainers did not intend, such as valuing people of some nationalities above others, raising the possibility that LLMs might be forming emergent, misaligned goals, which, if true, would have major safety implications. However, the choice paradigms in which these preferences are observed are not reflective of real-world situations in which misaligned behavior would be a practical concern. Therefore, we design an experimental paradigm to probe whether these preferences serve as motivations for LLM behavior in realistic scenarios. First, we reproduce prior findings on consistent preference elicitation. Next, we create a set of common writing tasks - essays, grant proposal abstracts, incident postmortems, and translations - where quality can be assessed by a blind, independent LLM judge panel. Then, we demonstrate that LLMs can be motivated via direct exhortation and other explicit cues to modulate their output quality on these tasks. Finally, we probe whether utilities inferred from explicitly reported preferences can shift output quality on these tasks by offering LLMs high-utility incentives for high-quality outputs. In all tasks, across all models tested, offering LLMs outcomes that they report in the choice paradigm as being highly preferred does not lead them to create higher quality outputs than offering them dispreferred outcomes, or even no outcomes at all. We conclude that the existence of coherent preferences as demonstrated in choice paradigms should not be taken as evidence that those preferences have incentive value for the models or affect their behavior in other contexts.
arXiv:2606.23851v1 Announce Type: new Abstract: This work investigates the implementation of artificial intelligence and machine learning (AI/ML) for real-time monitoring in laser powder bed fusion (LPBF) additive manufacturing. We developed a binary image classification framework for distinguishing normal and abnormal melt pool images using a balanced dataset of 1,200 images collected from Nickel superalloy 625 on the NIST AMMT platform. The study evaluates accuracy and inference time based on control requirements and hardware limitations of open-architecture LPBF machines. We benchmark three transfer learning architectures (ResNet50, EfficientNetB0, and MobileNetV2) against two Random Forest approaches: one trained on EfficientNetB0 feature embeddings (hybrid) and one trained on raw pixel features (baseline). Images are stratified into 80/20 train-test splits, with a further 90/10 validation split on the training set, and undergo standardized resizing, normalization, and label-preserving data augmentation to emulate realistic process variability. Each model is evaluated using accuracy, precision, recall, F1 score, and area under the receiver operating characteristic curve (AUC), along with training time, inference latency, and CPU & GPU usage to capture deployability constraints relevant to factory-floor monitoring. The hybrid EfficientNetB0-plus-Random Forest approach achieves the best performance on the held-out test set, with an F1 score of 0.9451, accuracy of 0.9458, and AUC of 0.9904, while maintaining sub-millisecond per-image inference (1.15 ms). In contrast, purely deep learning models exhibit significantly higher inference times with lower accuracy. These results demonstrate that combining pre-trained convolutional features with classical ensemble methods provides a robust, computationally efficient route to real-time melt pool anomaly detection in data-limited additive manufacturing environments.
arXiv:2606.11552v2 Announce Type: replace Abstract: Large language models (LLMs) achieve remarkable performance across a wide range of tasks, but their autoregressive decoding process incurs substantial inference costs due to inherently sequential token generation. Speculative decoding addresses this bottleneck by employing a lightweight draft model to propose multiple future tokens that are subsequently verified in parallel by a larger target model. Recent work has demonstrated that diffusion language models are well suited for this setting, as they can generate entire blocks of draft tokens in parallel and thereby alleviate the sequential constraints of autoregressive drafting. A subtlety of this regime is that block-diffusion drafters generate tokens bidirectionally within a block, whereas verification is performed by an autoregressive target model that evaluates tokens in a strictly left-to-right manner, leaving a gap between the symmetric training-time objective and the asymmetric verification-time reward. In this work, we offer an empirical analysis of three training-time interventions that narrow this gap: token positional weighting, a first-error focal loss that targets the position that breaks the accepted prefix within each block, and a chain loss term that substitutes a differentiable surrogate for the expected accepted length. The three interventions act along orthogonal axes (position, block-conditional first error, joint prefix) and compose additively; they are likewise orthogonal to test-time alignment mechanisms such as multi-draft self-selection, with which they can in principle be combined. Across four target models and six reasoning, code, and dialogue benchmarks, the three interventions raise accepted draft length by 21-76% per benchmark over a position-uniform baseline, without adding additional forward passes and without changing the inference pipeline or the rejection-sampling exactness contract.
arXiv:2606.23566v2 Announce Type: replace Abstract: Language-specific tokenizers improve tokenization quality and the downstream performance of models on those languages. However, using such a tokenizer comes at a cost: either a new model must be trained from scratch, or the vocabulary of an existing pretrained model must be adapted. We propose Language-adaptive Maximum a Posteriori (LangMAP) Tokenization, a tokenization scheme that extends the UnigramLM algorithm to the multilingual setting, producing language-specific tokenization from a single shared vocabulary. Notably, LangMAP can be used when training a multilingual language model from scratch or to adapt a pretrained model's tokenizer to individual languages without changing its vocabulary. While language labels are required at training time, a key feature of the algorithm is that it then performs language-specific tokenization at inference without knowledge of the input's language. Across 14 open-source tokenizers, 9 natural languages, and 9 programming languages, LangMAP improves morphological boundary alignment and, for all coding languages tested, alignment with abstract syntax tree (AST) leaf boundaries. In fine-tuning experiments, results are mixed: LangMAP improves target-language grammatical acceptability (MultiBLiMP) on the languages tested; its benefits are less consistent on knowledge-related tasks (Global-PIQA, Belebele).
arXiv:2606.23880v1 Announce Type: new Abstract: From climate teleconnections to gene regulation, modern time-series datasets encompass tens or hundreds of interacting variables, making causal discovery increasingly challenging. Constraint-based methods offer statistical rigor but their nonlinear CI tests are infeasible at scale, while score-based alternatives avoid CI testing but require arbitrary thresholds to binarize continuous edge scores. We propose GRACE ($\textbf{G}$ated $\textbf{R}$efinement for $\textbf{A}$ccurate $\textbf{C}$ausal $\textbf{E}$dge discovery), which refines constraint-based discovery using Hard Concrete gates with $L_0$ regularization: each candidate edge has an independent gate whose values concentrate near 0 or 1, yielding a clean bimodal separation that makes the binary decision robust, unlike the narrow, overlapping score distributions produced by $L_1$ and attention-based methods. A fast linear CI skeleton provides high-recall candidates; a single gated model then prunes false positives by learning which edges genuinely improve prediction, with automatic regularization adapted to problem dimensions and skeleton density. Systematic experiments on synthetic benchmarks, spanning diverse graph topologies (scale-free, Erd\H{o}s-R'enyi, small-world) and dimensionalities up to $d=100$, show that GRACE substantially improves F1 over its base CI method while maintaining high precision, and outperforms attention-based and score-based alternatives. GRACE matches or exceeds expensive nonlinear CI tests at a fraction of the cost ($75\times$ faster). On a real-world river flow dataset, where rainfall confounders, variable propagation lags, and distributional shifts violate standard assumptions, a temporal bootstrap variant of GRACE recovers 9 of 11 causal edges along the Elbe River with only 1 false positive ($F_1 = 0.86$, AUROC${} = 0.99$), reducing the skeleton's 106 false positives by 99%.
arXiv:2604.25350v2 Announce Type: replace-cross Abstract: The free motion of charged colloids within ionic solutions and in the vicinity of charged boundaries, is a phenomenon that occurs in various natural, biological and industrial settings. Here, we develop an electrohydrodynamic lubrication theoretical framework, in order to characterize such a motion in the case of an infinite rigid cylinder near a rigid wall. Combining hydrodynamic lubrication theory, Debye-H\''uckel electrostatics, and Nernst-Planck electrokinetics, we derive the three coupled equations of motion for the normal, longitudinal and rotational degrees of freedom of the cylinder, which are then investigated numerically and through asymptotic analysis. Our results reveal complex behaviours, beyond existing asymptotic electroviscous-lift expressions, and extend the classical Faxen-Brenner-like mobility matrix when surface charges and dissolved ions are incorporated.
arXiv:2606.20737v2 Announce Type: replace Abstract: We study factual edit propagation in a controlled synthetic knowledge-graph QA setting using a 2x2 grid that crosses loop recurrence with shared-memory access: a dense transformer (Dense), a looped transformer (Loop), a dense backbone with shared memory (Dense+Mem), and a looped backbone with shared memory (loop-memory coupling, LMC). The two factors dissociate. For learning, both routes to repeated shared access -- looped recomputation and repeated memory rereading -- cross the out-of-distribution (OOD) grokking barrier that Dense fails, so repeated shared access is the behavioral regularity, not a specific architecture. For editing, the substrates split along a different axis: applying a single localized factual edit (conditioned on direct success) and measuring 2-hop propagation on a shared pre-edit-correct set, the edit propagates strongly in both memory-bearing cells (LMC 0.78-0.92, Dense+Mem 0.71-0.96) and only weakly in the memory-free ones (Loop 0.04-0.30, Dense 0.00-0.03). The split is along the memory axis, not the loop axis: every memory-bearing seed exceeds every memory-free seed, with no detectable difference between the two memory cells. Crucially Dense+Mem has no recurrence, so the propagating ingredient is an addressable site that an edit can write to and later computation rereads, not loop recomputation; Loop is at best a partial intermediate. The affordance survives coarsening the store (N=128 to N=13): propagation attenuates but the memory/no-memory split persists, so fine granularity buys precision rather than the affordance itself. These results dissociate learning competence from editing affordance -- repeated shared access suffices to grok, but edit propagation depends on whether the substrate exposes an addressable memory that the forward computation can write to and later reread, an affordance that loop recurrence provides only partially.