Papers

We present a method for constructing a simple quadrangulation of the sphere from an unordered set of $n$ direction vectors, with application to parametric volumetric models of skeletal joints. We first select a quadrangulation encoding the joint topology by exhaustively enumerating candidate dual graphs, prioritizing convexity and regularity of the resulting quads. We then realize this topology on the sphere without crossings by solving a conflict-constrained assignment problem with lazy cyclic-order separation, and recover the corresponding primal graph. Experiments on direction sets from spherical Fibonacci sampling of size $3$ to $20$ show that the method robustly produces valid, high-quality quadrangulations across a range of input configurations. The resulting primal graph is directly converted into exact rational Bézier shell and column volumes, yielding a compact and precise volumetric representation of the joint. The approach extends naturally to multi-joint assemblies, enabling the construction of truss-based lattice structures within 3D domains.

Robot manipulation has traditionally relied on end-effector-centric interactions. In contrast, humans routinely exploit whole-arm contacts to improve efficiency, robustness, and adaptability, especially when the hands are occupied or when precise grasping is unnecessary. Enabling similar capabilities in robots remains challenging: the robot must select where on its arm to make contact, generate effective non-prehensile interactions, and satisfy task-specific pose or safety constraints on the end-effector. We propose Whole-Arm Contact Learning (WACL), a hierarchical framework that describes each whole-arm contact task with three robot-agnostic representations and hands them to a quasi-static surface-contact optimizer: semantic keypoints that summarize the scene, a reference contact trajectory that specifies where and how the arm presses on the object, and safety-aware path constraints that a vision-language model generates from the task description to protect the end-effector's task. The robot's kinematic chain enters only in this optimizer, which grounds these representations onto the arm by jointly searching over which link makes contact, where on that link, and the joint configuration that realizes it. This decomposition lets WACL learn the contact trajectory from 10 bare-hand human demonstrations per task collected without teleoperation, while new end-effector constraints change only the optimizer's input rather than any trained component. Real-world experiments on a 7-DoF arm across four whole-arm tasks show that WACL sustains contact under end-effector constraints where a strong link-pose baseline fails (78% vs. 33% success) and generalizes to new constraints without retraining. Website: https://wacl-robot.github.io

In this paper, we present the first results for the one-loop decay $A \rightarrow ZZ\gamma$ in the Two-Higgs-Doublet Model and its extension with Vector-Like Quarks. We calculate the one-loop partial decay widths and branching ratios for both unpolarized and polarized final-state $Z$ bosons. The effects of Vector-Like Quarks propagating in the loop are also examined in our phenomenological analysis. We find that the branching ratios can reach $10^{-5}-10^{-4}$ for several benchmark points within the viable parameter space of the considered models. The contributions from Vector-Like Quarks can modify the total rates by approximately $5\%$. With the high luminosities expected at future multi-TeV muon colliders, these effects may be experimentally accessible.

Multi-modal semantic segmentation adds an auxiliary sensing stream to RGB images (RGB-X), such as thermal, LiDAR, depth, event, polarization, or light field. This can improve robustness in poor lighting and under motion blur. But different sensors often produce features that do not match well, and fusion can become unstable. To alleviate these issues, we propose a bidirectional polarity-aware cross-modality fusion (BPCF) module that effectively captures complementary cues while enhancing feature alignment. We evaluate the framework on six modality pairings: RGB-Thermal, RGB-LiDAR, RGB-Depth, RGB-Event, RGB-Polarization, and RGB-Light Field. Our model achieve state-of-the-art (SOTA) results on eight public datasets, including MFNet, KITTI-360, DELIVER, DDD17, DSEC, MCubeS, ZJU, and UrbanLF. The codes and trained models will be released upon acceptance.

Online reinforcement learning (RL) enables driving policies to adapt to changing deployment conditions, but requires safe exploration throughout training. Safety filtering alone does not ensure stable policy improvement or reduced dependence on corrections. We propose Safe Online Adaptive Refinement (SOAR-Flow), a framework for safe and stable online fine-tuning of offline-initialized Flow-based driving policies. A control barrier function (CBF)–model predictive control (MPC) filter corrects potentially unsafe actions before execution. To support stable policy updates, we introduce Gaussian-return-aware value calibration based on calibrated Q-learning (Cal-QL), using recorded behavior returns and predicted return variability to smooth conservative critic updates and mitigate policy degradation. Paired corrective learning teaches feasible action adjustments, while an intervention-to-go predictor encourages decisions requiring fewer future corrections. Its predictions also regulate guidance strength, allowing recovery supervision to weaken as predicted dependence decreases. Together with gradually reduced calibration regularization, this supports continued task improvement while preserving hard execution constraints. We analyze conditions for safe execution at control steps and diminishing auxiliary learning influences. We evaluate SOAR-Flow in simulation and on an uncrewed ground vehicle (UGV). The results demonstrate improved driving performance and reduced reliance on safety corrections during protected online fine-tuning. The demonstration video and project overview are available at https://lzqw.github.io/SOAR-Flow/.

For a Bruhat interval $[x,y]$ in an arbitrary Coxeter group which is a lattice, we give a type-uniform construction of a convex polytope whose complete face lattice realizes $[x,y]$. This gives a polytopal realization of the coincidence between classical Kazhdan–Lusztig polynomials for lattice Bruhat intervals and Stanley's toric $g$-polynomials. Consequently, the Kazhdan–Lusztig polynomial of a lattice Bruhat interval cannot have internal zeros, satisfies a multiplicative strengthening of monotonicity, and obeys the generalized upper bound theorem for polytopes. For lattice Bruhat intervals in finite Weyl groups, we explain the relation with the toric geometry of Richardson varieties and derive formulas for their local and global intersection cohomology, as well as a Boolean criterion for rational smoothness. We conjecture that the coefficients of these Kazhdan–Lusztig polynomials form an $M$-sequence.

Based on our previous results regarding critical exponent estimates for the 3D Ising model and scaling relations [1], we have proven that the exact value of the correlation-length exponent is $\frac{\sqrt[3]{2}}{2}$. This is in very close agreement with the most precise numerical estimates currently available. Using a group theoretical approach and some propositions of constructive geometry, we derived an exact construction of algebraic expressions for the critical exponents $\gamma$ and $\beta$. We proved that these expressions depend solely on the exact value of $\nu\,$.

Compared with a scalar Bose-Einstein condensate (BEC), a spinor BEC supports richer phenomena because of its internal spin degrees of freedom, including topological solitons such as domain walls or skyrmions. Here, we theoretically reveal three distinct axisymmetric soliton structures formed as a circular domain wall in a spin-1 BEC, connected within fixed winding numbers. We refer to these structures as the broken-axisymmetry (BA)-skin, antiferromagnetic (AF)-skin and normal-core phases. Their occurrence is controlled by the two dimensionless parameters in the absence of linear Zeeman coefficient $p=0$: $q/\mu$, the ratio of the quadratic Zeeman coefficient to the chemical potential, and $c_s/c_n$, the ratio of spin interaction to the density interaction. The BA-skin and AF-skin phases exhibit local easy-plane magnetic order and local easy-plane spin-nematic order, respectively, while the normal-core phase is similar to a doubly quantized vortex analogous to that of a scalar BEC with a density depletion at the center. We obtain semi-analytical estimates of selected transition boundaries separating these phases by analyzing the Bogoliubov-de Gennes equation and also numerically confirm them by solving the Gross-Pitaevskii equation. We estimate the skyrmion radius using a hydrostatic approximation and compare the result with numerical solutions. We envision our results on skyrmions in a spin-1 BEC can trigger further investigation on novel soliton structures in a magnetic superfluid.

Large language model (LLM) serving systems use token-level continuous batching to improve resource utilization, but heterogeneous response lengths complicate request scheduling. Prediction-guided short-first scheduling reduces average latency but can severely delay long requests, while unrestricted aging may over-promote delayed long requests under congestion. We propose Quota-Based Aging (QBA), a lightweight policy that combines predicted short-first admission with quota-bounded, debt-enforced promotion of aged predicted-long requests. We also use a reduced M/G/n queueing surrogate, calibrated from simulator traces, to explain first-order mean-latency trends. QBA is evaluated using a serving-aware event-driven simulator that models decode-priority batching, chunked prefill, token-budget constraints, and KV-cache-safe admission. We compare QBA with FCFS, predicted short-first, and aggressive aging under asymmetric prediction errors on synthetic and production-derived workloads. The evaluation considers queueing delay, throughput, utilization, and class-conditional tail performance. Results show that QBA provides a controllable middle ground between aggressive short-first prioritization and unrestricted aging, substantially reducing long-request tail latency while retaining much of the mean-latency benefit of prediction-guided scheduling.

Contextual forecasting seeks to improve time series forecasting by incorporating forward-looking information, such as planned events, external conditions, and expert expectations. However, reliable evaluation remains difficult: existing benchmarks often have limited domain coverage, lack standardized history-future splits, or provide text that is not guaranteed to inform the future. We introduce GIFT-Eval-CTX, a benchmark that isolates whether forecasting methods can translate verified future information into accurate numerical trajectories. Built upon GIFT-Eval, GIFT-Eval-CTX contains 10,000 instances across 20 datasets. Each instance provides structured effects derived from the realized future: Macro effects describe horizon-wide behavior, while Micro effects specify localized changes. We construct these annotations with an agentic pipeline combining LLM-based discovery and deterministic code validation. We benchmark TSFMs, frontier LLMs, and tool-using agentic systems. The results show that structured effects reduce aggregate relative MASE by 40.7% for GPT-5.6 Sol and 29.6% for Claude Opus 4.8 relative to history-only prompting. The evaluated MoiraiAgent system achieves the best overall result. We further analyze the contributions of structured effects and agentic systems to performance, and identify opportunities for future research.

This study evaluates the AutoSTEM workshop, an intervention for secondary education that combines a serious game with PID-controlled hardware. Seven workshops were conducted across Sweden and Italy. The six Swedish workshops adopted an explicitly SDT-informed facilitation framework, while the Italian workshop followed the existing educational robotics format without a formally defined SDT facilitation protocol. Self-reports, an objective knowledge quiz, and qualitative responses were analysed using multiple site-adjusted statistical approaches. Technical-specific self-efficacy increased substantially in both conditions, with large effect sizes, while objective knowledge showed only modest, non-significant gains. No significant site-adjusted differences were found between SDT and control conditions for the primary outcomes. Perceived engagement was higher in the control condition, although this finding should be interpreted cautiously. The objective knowledge composite remained positive but non-significant across sensitivity analyses. Female participants reported smaller PID self-efficacy gains than males. Qualitative findings supported evidence of PID learning while identifying technical and pacing barriers. Overall, the results suggest that the technology-rich intervention supports competence development, regardless of SDT-informed facilitation.

GUI grounding enables web agents to translate natural-language instructions into actionable interface locations. However, dense layouts and small, visually similar controls can lead models toward plausible but incorrect elements, while refinement around a single prediction may leave alternative targets unexplored. We propose the Multi-Hypothesis Verification Framework (MHVF) for training-free GUI grounding. It combines multi-hypothesis exploration with local visual verification using a single pretrained vision-language model. To explore alternative targets, MHVF constructs spatial hypotheses from multiple full-screen predictions and examines the corresponding regions through zoomed re-localization. To resolve ambiguity among the resulting candidates, it verifies their agreement with the instruction in marked local views. Verification scores then guide a second round of local search and rank candidates retained from both search levels for final selection. Experiments with five general-purpose and GUI-specific models across eight benchmark settings demonstrate improvements in grounding accuracy in most settings. MHVF improves UI-Venus-Ground-7B by 4.3 and 2.1 percentage points on Mind2Web and VisualWebBench, respectively, and raises Qwen2.5-VL-7B's accuracy on ScreenSpot-Pro from 23.6% to 53.6%.

Understanding facial behaviour in video requires answering three fundamental questions: when a key facial expression occurs, where the model attends on the face at that moment, and why that evidence is necessary rather than merely correlated. Today's multimodal large language models (MLLMs) can label and describe video, but they rarely provide faithful explanations, especially for small facial movements that change over time. We present VideoFace-X-LLM, a framework for explainable facial video understanding comprising two key components. First, we introduce ExplainFace, a weakly supervised video-text corpus for facial behaviour understanding, automatically constructed from CelebV-HQ videos and validated through human evaluation and a two-model audit (at most 2.3% contradictions), demonstrating consistent gains across two backbone architectures. Second, H-TCC (Hierarchical Temporal Spatial Causal Concepts), which finds the key temporal moments (When), the facial regions (Where), and verifies its causal necessity (Why) through targeted occlusion. Fine-tuning on ExplainFace raises appearance Macro F1 from 19.48% to 32.87% (+13.39), with gains of +4.38 on actions, +6.77 on temporal emotion, and +11.84 F1 on deepfake detection. We also measure the explanations pointing lift, face/edge lift, stability, an Adebayo sanity check, and a necessity test rather than only showing examples. Our models, code, and corpus are available on request for research use.

Level crossings, where two quantum energy levels intersect, are ubiquitous in quantum systems and play a central role in controlling quantum dynamics. In this study, we establish a framework that connects these level crossing structures to the ultimate precision limits of quantum metrology. By analyzing a Hamiltonian with the level crossings structure, we identify the conditions under which multiple parameters can be estimated simultaneously with maximal sensitivity near the level crossing point. We then construct a resource efficient sensing protocol based on unitary quench dynamics and Bayesian estimation, enabling high precision, low cost parameter estimation without requiring entanglement or complex control. We present a protocol whose estimation precision saturates the Holevo Cramer Rao bound, the tightest achievable precision bound in multi parameter estimation. Using the two level model and the two spin Ising model as illustrative examples, we demonstrate how the protocol can be applied to a system with level crossing Hamiltonian structure. Our findings reveal that level crossings serve as universal operating points for precision quantum metrology, offering a simple and scalable path toward practical multi parameter sensing across diverse quantum platforms.

A long-horizon manipulation task consists of primitive skills and the transitions connecting them. Learning skills that can be executed in sequence requires supervision for both, which is costly to collect on a robot. Human videos offer a scalable source of demonstrations, but transferring them to robots requires extracting robot-compatible observations and actions together with their skill structure. We present DexStitch, a framework that cuts long-horizon human videos into skills and transitions, converts them into aligned robot observation-action-skill data, and stitches the learned skills into a requested sequence. A skill classifier first segments the videos into manipulation skills and transitions. DexStitch then converts the human demonstrations into robot observations and actions while retaining their corresponding skill labels. During conversion, per-finger human-object contacts guide dexterous hand retargeting and robotic inpainting, preserving contact geometry and finger-object occlusion ordering. To stitch the learned skills into a sequence, a shared skill-conditioned policy is pretrained on the resulting synthetic data and finetuned on a mixture of synthetic data and a small set of single-skill teleoperation episodes. Conditioning on the current and next skills guides each transition toward the following skill, while predicted skill progression enables closed-loop switching. On a held-out session, the classifier agrees with manual annotations on 92.4% of frames. Across four household skills, DexStitch achieves 80.0% single-skill success and 77.5% sequence success.

Distributional critics model the full return distribution, providing richer signals for exploration and risk-sensitive control. Recent work has shown that flow-based critics can model these distributions flexibly and provide effective supervision across flow time for temporal-difference learning. However, unlike standard flow matching with fixed data endpoints, a flow critic is trained on Bellman targets recursively generated by a learned target critic under the current policy. Existing objectives therefore either use noisy single-endpoint supervision or rely on the learned flow itself for additional guidance. Our key observation is that an intermediate bridge state may be compatible with several Bellman endpoints, so its velocity should average their directions under the corresponding posterior. We propose Posterior-Averaged Bellman Flow Matching (PA-BFM), which samples an endpoint pool, infers the closed-form posterior over the latent anchor, and regresses to the posterior-mean velocity. Uniform anchor selection makes this target an exact finite-pool Rao–Blackwellization of the single-endpoint target: for every \(K\), it preserves the population target, cannot increase conditional variance, and its conditional variance scales as \(O(K^{-1})\). The endpoint pool supports conservative updates, while the twin return flows provide critic-disagreement and unsupported-mass diagnostics. Controlled studies show that PA-BFM reduces fixed-query target mean-squared error by \(79.7\%\) and target-induced gradient variance by up to \(98.0\%\), while improving return-distribution fidelity. Across 50 OGBench tasks and four offline-to-online scenarios, PA-BFM achieves higher aggregate performance than Value Flows.

We calculate analytic expression for the cross section up to order $(eB)^2$ for the production of a heavy quark $(Q)$ of mass $m_Q$ through the gluon $(g)$ fusion process $g+g\to Q+\bar{Q}$ in the presence of a constant weak external magnetic field $(B)$ with strength $eB \ll m_Q^2$. This is done using the weak-field expansion of the fermion propagator calculated within Schwinger's proper-time formalism. Owing to the hierarchy of scales ($m_Q^2 \gg eB$), we neglect the effects of the magnetic field on the quark and antiquark spinors in the external lines. Since the magnetic field breaks rotational invariance, the cross section becomes anisotropic and depends on the relative orientation angle $(\theta)$ between the magnetic field and the incoming gluons in the center-of-mass frame. The lowest-order magnetic correction to the vacuum cross section scales as $(eB)^2$ and is found to have $\cos 2\theta$ and $\cos 4\theta$ dependence, reflecting the breaking of rotational symmetry by the magnetic field. The cross section is maximized when the gluons propagate along the field direction ($\theta = 0, \pi$) and minimized in the perpendicular configuration ($\theta = \pi/2$). The resulting asymmetry peaks near the production threshold and falls with increasing center-of-mass energy, scaling as $(eB)^2/m_Q^4$ with the quark mass. These results provide direct, calculable signatures of the initial-stage magnetic field in relativistic heavy-ion collisions which can, in principle, be imprinted on heavy-flavour observables measured in the final state of heavy-ion collisions.

We study industrial fault diagnosis as a classification task that assigns sensor time-series windows to predefined health states or fault types. Scarce labels and changing operating conditions make repeated training of specialized classifiers costly. Yet structures such as periodicity, trends, and transients can remain diagnostically informative and be examined through reusable analytical tools. LLM agents can select and interpret these tools, but fixed rules or training-derived guidance may not resolve each signal's ambiguity, while unconstrained interpretation may overstate weak or conflicting results. We introduce Agentic4PDM, a framework that separates adaptive investigation from control over evidence use. LLM agents compare each signal with labeled references from every candidate class, identify unresolved distinctions, and propose follow-up measurements. A deterministic software controller, called the host, enforces contract-checked gated adjudication (CCGA): it checks requests, requires commitment to a complete tool batch before results are observed, and assigns evidence roles based on tool checks against labeled references. These roles distinguish corroborative support from advisory information. The Judge Agent proposes a diagnosis, while the controller checks protocol compliance and records the investigation. Agentic4PDM supports training-free diagnosis using pretrained representations and labeled references without task-specific diagnostic models, and a standard setting that additionally investigates trained-model predictions. Experiments on selected bearing and motor diagnosis challenges and an unfiltered industrial test set supplied by our customer show gains in aggregate macro F1 over the evaluated diagnostic and agentic baselines. Ablations further support the value of the governed workflow.

Hybrid beamforming (HBF) reduces the hardware cost of massive MIMO base stations, but its channel acquisition requires recovering the full spatial-domain channel from highly compressed beam-domain observations. This paper proposes a learned score-based turbo message passing (L-STMP) framework for uplink HBF channel recovery under extreme spatial compression. By deeply unfolding the turbo message-passing loop and introducing trainable calibration parameters and learnable score priors in each layer, L-STMP liberates STMP from the idealized compression-matrix requirements of right-rotational invariance, and reduces inference complexity to only several unfolded layers. In addition, to preserve dominant channel components that random discrete Fourier transform (DFT) selection may miss, we introduce precoding matrix indicator (PMI)-assisted beam selection as a low-overhead directional prior. Simulation results demonstrate that L-STMP consistently outperforms existing baselines and obtains substantial reconstruction gains under PMI-assisted compression in channel reconstruction.

For a $d$-dimensional Noetherian local ring $(R,\mathfrak{m})$, with $d>0$, let $c_{\infty}(R)$ be the limit, as $N\to\infty$, of the suprema of $e_R(I)/(d!\ell_R(R/I))$ over ideals $I$ with $\sqrt{I}=\mathfrak{m}$ and $\ell_R(R/I)>N$. We prove that if $R$ is complete, reduced, equidimensional, and Cohen–Macaulay on the punctured spectrum, with $d\geq2$ and infinite residue field, then every nonzerodivisor $x\in\mathfrak{m}$ satisfies \[ c_{\infty}(R)\leq \max\left\{c_{\infty}(R/xR), \sup_{\mathfrak{p}\ne\mathfrak{m}}e(R_{\mathfrak{p}})\right\}. \] As an application, we prove the asymptotic Lech conjecture of Huneke, Ma, Quy, and Smirnov in arbitrary characteristic: $c_{\infty}(R)=1$ whenever $\widehat{R}$ has isolated singularities. This also settles the conjecture of Ma and Smirnov that the Lech–Mumford constant $c_{\mathrm{LM}}(R)$ is attained when $\widehat{R}$ has isolated singularities and $c_{\mathrm{LM}}(R)>1$. For reduced equidimensional Cohen–Macaulay standard graded algebras of dimension $\geq2$ over infinite perfect fields, we prove attainment and rationality of $c_{\mathrm{LM}}$ for the completed vertex local rings of sufficiently high Veronese subrings. In characteristic zero, a refinement of our method produces nonisolated klt affine cones in every dimension $d\geq3$, with $d-2$-dimensional singular loci, whose completed vertex local rings have $c_{\infty}=1$ and arbitrarily large $c_{\mathrm{LM}}$.

We develop an algebraic theory of local intertwining periods for symmetric pairs $(G,H,\theta)$ over non-archimedean local fields $F$ of characteristic different from $2$, and finite-length representations of Levi subgroups of $G$ with coefficients in an algebraically closed field $R$ of characteristic different from the residual characteristic of $F$. For open orbits this algebraic version is essentially contained in the work of Blanc and Delorme. The algebraic theory relies on a crucial new uniqueness result, extending a result of Blanc and Delorme for open periods, which asserts that an admissible intertwining period is the only $H$-equivariant linear form extending the corresponding integral on the space of well-supported sections. Moreover the algebraic construction does not require the unimodularity hypothesis imposed in the work of Matringe–Offen–Yang. When $R=\mathbb{C}$, we also remove this assumption in the classical non-archimedean analytic setting, and prove convergence of intertwining periods for all symmetric spaces. At the same time we extend the context of the whole theory in the following manner: we allow $F$ to be of positive odd characteristic, we allow twisting characters in the periods, and we allow $H$ to lie between $(\mathbb{G}^\theta)^\circ(F)$ and $G^\theta$. Finally, in the algebraic setting, as in the work of Blanc and Delorme on open periods, we obtain explicit bounding polynomials for the singularities of the intertwining period. Such bounds are a key ingredient of a new proof of the cuspidal case of a conjecture of D. Prasad on quaternionic symplectic models.

Forced aligners locate each word of a known transcript in the audio, and many speech pipelines depend on them. In our own data, a widely used CTC aligner inserts a pause at nearly every word boundary, and we found almost no aligner evaluated on telephone audio. We compare nine aligner configurations against human word boundaries on read, conversational and telephone English. The Montreal Forced Aligner (MFA) is the most precise, while CTC aligners consistently start words late, because most frames inside a word go to the blank token. We then fine-tune Whisper with LoRA to predict each word's start and end as timestamp tokens, training only on boundaries placed by people: 47 hours in 87 languages. The resulting aligner is more precise than MFA on every English test condition, and than MFA's own Korean and Dutch models. With FA-Bench's protocol and code on TIMIT, its word-boundary error is 11.6 ms, against 18.2 ms for the best published system. Two choices matter most on telephone conversations: placing the timestamp slots after the word they time, and training on noisy transcripts, which also makes the aligner nearly insensitive to errors in the transcript. The released model reaches 16.0 ms on real landline calls, against 36.5 ms for MFA. The models are available at https://huggingface.co/collections/Scicom-intl/force-alignment

Replacing the circle by a manifold leads to current Lie algebras and their central and derivation extensions. We separate the coefficient algebra, harmonic analysis and geometric data. Differentials modulo exact forms describe central extensions; divergence-free vector fields give explicit cocycles, whose compatible derivations satisfy a stabilizer equation. Compact examples include the three-sphere and its current–Virasoro extension. We also explain the analytic qualifications for SL(2,R) and its symmetric space and the role of transported multiplication in soft geometries. The recent harmonic constructions of Campoamor-Stursberg, Marrani and Rausch de Traubenberg are compared with the cohomological work of Coquereaux, Frappat, Ragoucy and Sorba and Castellani's loop geometry and infinite-dimensional free differential algebras. A coefficient-algebra construction connects currents to higher brackets and clarifies the additional requirements of a dynamical theory. For Lie algebras of order F, central extensions obey coupled binary and symmetric F-ary conditions, with explicit examples and obstructions.

We introduce Diffuplex, a full-duplex spoken language model based on discrete diffusion that achieves an unprecedented balance between linguistic intelligence and response efficiency. With thinking mode, Diffuplex delivers 13.0 tokens per forward pass, reaching an average of 1872 tokens per second and substantially surpassing the frontier autoregressive (AR) models in decoding speed. Without thinking, Diffuplex matches the first-audio latency of existing full-duplex models while achieving markedly stronger linguistic intelligence. This capability is enabled by separating when to say from what to say at the training level, rather than compressing both into a single autoregressive timeline. During inference, Diffuplex continuously tracks the user speech through an AR interaction policy, while independently generating and refining response content through parallel diffusion. Consequently, semantic computation is no longer bounded by the perception audio frame rate, allowing intelligence and responsiveness to improve simultaneously. Taken together, these results establish a new paradigm for spoken agents that can sustain real-time dialogue while devoting parallel computation to intelligent response generation. Training code and checkpoints will be avaliable soon.

The Kolmogorov-Arnold Network (KAN) challenges traditional MLPs by placing learnable activation functions on network edges. However, scaling KANs to spatial data exposes limitations in default B-Splines. While hybrid architectures mix multiple basis families on a single edge to resolve this, monolithic superposition causes geometric overlap, parameter explosion, and severe numerical ill-conditioning ($\kappa(G) \approx 10^{16}$).To overcome this, we propose Parallel BesselKAN, which isolates B-Spline and low-order Bessel branches into parallel processing pipelines integrated via a parameter-free quadratic interaction. This decoupled architecture acts as an efficient internal ensemble, bypassing monolithic stability bottlenecks. It compresses the parameter footprint by $35\%$ while achieving $97.99\%$ accuracy on MNIST. Although standard MLPs retain a hardware speed advantage for pure vision tasks, Parallel BesselKAN provides a robust, parameter-efficient framework stabilizing mixed-basis networks and paving the way for scalable spatial adaptations like Convolutional KANs.

Research on gender-related online content is often operationalized through harm-centered tasks that collapse non-harmful gender-related and unrelated content, while existing datasets remain concentrated in high-resource languages and Global North settings. To bridge these gaps, we introduce GENDIS, a dataset of 9,600 multilingual, human-annotated YouTube comments from India, Kenya, and Nigeria (3,200 per country), annotated using a two-level framework that first identifies whether gender is referenced and then classifies gender-related content as negative, positive, or neutral. This separation provides a denominator for gender-related discourse independent of harm, enabling both prevalence estimation and characterization of its evaluative range. In the uniform random stratum, 23.8-35.0% of comments are gender-related, and among those with a resolved stance, 82% are neutral or positive. Across models, gender-reference detection is substantially easier than stance classification, especially for positive and neutral speech. Lightly supervised classifiers match frontier zero-shot large language models at substantially lower cost and outperform generic toxicity detection, while cross-country transfer remains limited. GENDIS provides a locally grounded resource for measuring when gender is invoked online and how that reference is framed.

Theory of mind (ToM), the ability to reason about others' mental states, underlies effective communication and coordination. Its higher-order form enables reasoning over nested perspectives in strategic interactions. However, existing evaluations often equate belief-nesting depth with reasoning order, allowing nominally higher-order questions to admit lower-order shortcuts. We formalize effective order through the distinct events supporting successive belief prefixes, revealing widespread order inflation across existing benchmarks. We introduce HopToM, with certified full-order questions and controls for lower-order alternatives. Under direct prompting, mean accuracy across four frontier LLMs falls from 73.5% at first order to 20.0% at fourth order, and human participants also struggle. Causal interventions link a recurring class of errors to evidence for shallower belief prefixes. We introduce HopTrace, a plug-and-play inference-time method that encodes each interaction once into perspective records and resolves nested queries deterministically. On certified narrative questions, it improves fourth- and fifth-order accuracy by 22.0 and 38.5 percentage points over strong prompting baselines, while using 20-39 times fewer tokens than multi-call methods. These results sharpen how higher-order ToM is evaluated and offer a practical way to improve perspective reasoning in interactive agents.

In this paper, we introduce TempFit, a novel approach for the problem of Panoptic Symbol Spotting on CAD drawings. Existing methods treat symbol spotting as a per-primitive classification or mask-prediction task, with no explicit constraint enforcing that the primitives assigned to one instance form a spatially connected group. As a result, predictions may erroneously merge primitives from distant symbols into a single instance or fragment a coherent symbol across multiple predictions. Motivated by the professional "retrieve-and-transform" CAD design workflow, we reformulate this task as a Template Fitting problem, where each foreground instance is detected by retrieving and aligning a canonical template from a library, enforcing structural completeness by construction. Building on this formulation, we present an autoregressive transformer that unifies template retrieval and pose prediction within a single end-to-end pipeline. The framework treats the template library as an explicit geometric memory, parsing background regions in parallel via learnable background tokens and synthesizing foreground instances autoregressively while attending to template-specific geometry at each step. Experiments on FloorplanCAD and our newly collected IndustryCAD dataset, a more challenging benchmark with complex industrial symbols, heavy occlusion, and wide scale variation, demonstrate state-of-the-art performance with particularly large gains on IndustryCAD. Project page: https://tempfit-project.github.io/.

We determine explicit bulk and worst-case total variation cutoff profiles for the configuration law of a two-species Togashi–Kaneko model with a fixed population of $n$ particles and weak mutations. Boundary initial conditions introduce an additional logarithmic growth phase and change both the leading mixing-time coefficient and the limiting cutoff profile. Let $\kappa_1,\kappa_2>0$ be the autocatalytic rate constants and put $\lambda_*=\kappa_1\kappa_2/(\kappa_1+\kappa_2)$. For deterministic initial configurations whose density converges to $m\in(0,1)$, the total variation distance to equilibrium at time $t_n(s)=\frac{\log n}{2\lambda_*}+\frac{s}{\lambda_*}$ converges to $\mathcal G(r(m)e^{-s})$ for every fixed $s\in\mathbb R$, where $\mathcal G(r)=\mathrm{d}_{\mathsf{TV}}(\mathcal N(r,1),\mathcal N(0,1))$, $r(m)>0$ is explicit, and $\mathcal N(a,b^2)$ denotes the Gaussian distribution with mean $a$ and variance $b^2$. For the worst-case distance $\mathscr D_n(t)$ over all initial configurations, we prove $\mathscr D_n(T_n+s/\lambda_*)\to\mathcal P(s)$, where $T_n=\left(\frac{1}{\min\{\kappa_1,\kappa_2\}}+\frac{1}{2\lambda_*}\right)\log n$.The profile $\mathcal P$ is the total variation distance between the equilibrium Gaussian and a Gaussian mixture whose random displacement is determined by a gamma-distributed amplitude arising from early growth at the boundary. Unlike the bulk profile, $\mathcal P$ depends on the mutation parameters. Both cutoffs have windows of constant order, and we identify the asymptotically worst boundary configurations.

Neopentane-derived plastic crystals undergo order-disorder phase transitions accompanied by exceptionally large entropy changes, making them attractive solid-state materials for thermal energy storage and barocaloric cooling. Yet it remains unclear whether the microscopic nature of this entropy change is an intrinsic property of each compound or can be understood generally and tuned systematically through molecular design. Here, we use classical molecular dynamics simulations of neopentyl glycol (NPG), pentaglycerine (PG), and pentaerythritol (PE), validated against experimental transition temperatures, enthalpy changes, and phase diagrams, to show that the character of the phase-transition entropy undergoes a complete inversion as the hydrogen-bond connectivity of the molecule increases: orientational disordering dominates in NPG, whereas lattice vibrations dominate in PE and become remarkably robust against pressure. PG occupies a balanced intermediate regime in which the orientational, conformational, and vibrational channels contribute comparably. This inversion reflects two coupled microscopic mechanisms: a growing low-frequency gap between the phonon densities of states of the ordered and disordered phases as the hydrogen-bond network densifies, and an increase in molecular symmetry that progressively suppresses orientational disordering. Because the vibrational channel is substantially less sensitive to pressure than the orientational one, hydrogen-bond density emerges as a practical, chemically accessible design handle for tailoring not only the magnitude but also the pressure robustness of the caloric response in hydrogen-bonded plastic crystals, with direct implications for the search for new barocaloric and thermal-energy-storage materials.