Papers

When a team of robots delegates planning and mutual trust to LLM agents, a single compromised robot can corrupt the shared outcome. We study this threat in a grounded task: a multi-robot survey in which measurements can be verified against the physical world, but every verification costs budget that would otherwise advance the mission. We treat the compromised robot as a stealthy adversary in the system-theoretic sense: it is limited not by an energy bound but by the team's own detectors. We then derive two bounds. First, the probability that the adversary's reports are verified is bounded below in terms of the degrees in the communication graph and the verification budget. Second, the map error caused by any stealthy adversary is bounded above by the value of a linear program over the adversary's bias distributions; its solution is an exchange rate between stealth budget and damage: below a critical verification level the worst stealthy attack tells rare, full-magnitude lies on the records least likely to be verified, and above it the better purchase is small biases hidden in the noise. In experiments where the honest robots are LLM agents, both bounds hold at the budget the attack actually spent. The experiments also show that which records an LLM robot re-checks is unbiased, but how much it re-checks is unpredictable.

Accurate thyroid ultrasound segmentation is often challenged by low contrast, speckle noise, and unclear boundaries. Although recent methods have improved segmentation accuracy, many rely on resource-intensive architectures or lack explicit integration of multiscale features with global biomedical visual guidance. In this paper, we introduce ThyCLIPNet, a lightweight semantic-guided hybrid encoder-decoder framework that integrates BiomedCLIP-derived biomedical semantic guidance into a lightweight multi-scale CNN segmentation pipeline. The encoder integrates MobileNetV2 with efficient channel attention, while atrous spatial pyramid pooling and a custom convolutional block attention module enrich bottleneck features. The decoder combines hierarchical skip connections and lightweight attention refinement with a BiomedCLIP-guided gated fusion pathway that projects vision-only global biomedical embeddings into decoder feature space and selectively integrates them through semantic-local fusion and spatial gating. To the best of our knowledge, ThyCLIPNet is among the first lightweight thyroid ultrasound segmentation frameworks to use BiomedCLIP's vision encoder alone for image-only global semantic guidance without text prompting. Experiments on TG3K, TN3K, DDTI, and PKTN achieve dice similarity coefficients of 96.22%, 87.58%, 84.73%, and 80.70%; intersection over union scores of 92.72%, 77.91%, 73.51%, and 67.64%; and 95th-percentile hausdorff distances of 3.75, 16.38, 18.23, and 10.86, respectively. ThyCLIPNet uses 8.55M parameters and 22.99G FLOPs. Overall, the results support integrating global biomedical semantic guidance with lightweight multi-scale CNN representations for robust and computationally efficient thyroid ultrasound segmentation. Source code: https://github.com/Tasnim-Jahan/ThyCLIPNet. [Abstract shortened for arXiv. See PDF for full abstract.]

This paper illustrates the properties of a new CBF-governor that enables simultaneous input and output constraint satisfaction for a class of multi-input linear time-invariant systems with state feedback and integral action. This governor is designed so as to modify the reference command while preserving the nominal feedback controller. Necessary and sufficient conditions are derived in a companion paper under which the governor is shown to be feasible at every state of a prescribed operating set and ensures simultaneous input and output constraint satisfaction, forward invariance, and bounded closed-loop solutions. These theoretical results are illustrated in this paper through several numerical examples. In each of these examples, we show how the goals of the CBF-governor are met and corroborate the corresponding necessary and sufficient condition. The computational burden associated with the proposed CBF-governor is also articulated, with its online component comparable to either that of a QP solver or determined using an exact closed-form solution. The offline component is related to the checking of the SIOCF condition which is of $O(N)$.

Force-aware manipulation typically relies on specialized force or tactile sensors. We show that force-aware manipulation can instead be achieved through visual force prediction from the deformation of a compliant Fin Ray gripper. Our approach trains two models. First, we train a visual force estimator on calibration data and use it to annotate task demonstrations with force estimates. Second, we train an action–force proposal policy on these force-augmented demonstrations to jointly generate candidate robot actions and their associated forces. At test time, we sample candidate actions and the forces they are expected to produce, then execute the action whose predicted force is closest to a target from the demonstrations. We evaluate our approach on berry picking, empty-can grasping, in-hand reorientation, and plug insertion. Our results show that visual force prediction can guide inference-time action selection for contact-rich manipulation without requiring force or tactile sensors at deployment.

Early-stage computational drug discovery requires coordinating heterogeneous scientific tools across multi-step workflows. We present a lightweight, tool-augmented framework in which a locally deployable compact language model plans calls to 18 modular tools. A Unified Molecular Schema maintains shared molecular records, while a plug-in interface supports tool replacement and extension. We construct 1,263 manually refined query-plan pairs through workflow-graph path coverage and apply LoRA fine-tuning to three compact model families. Under the query-level split, all fine-tuned models generate fully parseable and schema-compliant plans on 47 held-out cross-group queries. Llama 3.2-3B achieves a tool-selection F1 of 0.998, sequence exact match of 0.979, and argument F1 of 0.960. Under the stricter workflow-grouped split, which excludes identical ordered tool sequences across partitions, sequence exact match reaches 0.452 to 0.548, highlighting the remaining difficulty for compact models in generating complete workflow paths unseen during training. These results demonstrate the feasibility of compact, locally deployable planning while identifying compositional generalization as an important direction for further improvement.

Counterexamples to the Eisenbud-Schreyer conjecture on the existence of Ulrich bundles have appeared recently (see Anghel [A1] and the third author [Lo]). As weakly Ulrich bundles are a useful generalization of Ulrich ones, in this paper, we classify smooth surfaces $S\subseteq \mathbb P^N$ with weakly Ulrich tangent bundle. They are Del-Pezzo surfaces of degree at least 5 and we classify the possible embeddings. An interesting feature is the appearance of two infinite series of polarizations, which is quite different from the Ulrich case [BMPT, LR]. We also achieve the classification of surfaces with initialized aCM tangent bundles and, as a consequence, of surfaces with almost Ulrich tangent bundles.

Ultraperipheral collisions of heavy ions, dominated by the electromagnetic interaction, allow to investigate the photoproduction of baryons and mesons from the excited nuclei. At LHCb, thanks to the SMOG2 system, these studies can also be conducted in fixed-target collisions, exploiting a variety of target gases ranging from $^{4}$He to Xe. In the present work, we adapted our photonuclear model and the Giesen-Boltzmann-Uehling-Uhlenbeck (GiBUU) generator to study photoproduction of protons, neutrons as well as pions and kaons from excited $^{40}$Ar nuclei in fixed-target ultraperipheral collisions with the LHC $^{208}$Pb beam. We present predictions for the total and differential photoproduction cross sections in $p$ and $p_T$, both in the full phase-space and within the geometrical acceptance of the LHCb detector.

Reasoning models can remain capable of solving a task while still defaulting to cheaper but misleading shortcuts. This creates a central oversight problem: when a model gives an answer with plausible but incomplete reasoning, can an overseer determine whether that output should be trusted? To study this problem, we introduce PyINE, a framework for scalable elicitation and oversight using instrumented Python programs as a verifiable execution substrate. In PyINE, programs define task environments, execution traces provide authoritative labels for outcomes and intermediate facts, and task variants can be generated mechanically rather than through static human annotation. We instantiate the framework in PyINE-v1, a first release built from nearly one million deterministic execution traces and over 500,000 matched LLM-generated code variants used for counterfactual evaluation. Using standard RL with verifiable rewards on cue-varied tasks, we train a shortcut-following model that improves substantially at predicting execution outcomes while still making systematic errors when misleading human-facing cues conflict with the program's realized behavior. We then evaluate activation probes, trained text classifiers, prompted judges, and a lightweight debate protocol as overseers of this model. We find that performance pooled at the dataset level can hide weak coverage of the failures that matter most: cheap learned overseers often miss rare shortcut-driven errors, while stronger model-based checks are more balanced but substantially costlier and harder to turn into reliable thresholded decisions. PyINE-v1 turns this failure-mode coverage problem into a reusable experimental setting for developing oversight methods that are verifiable, failure-mode-aware, and cost-sensitive.

A pedestrian with a phone could be shown where nearby people will be in the next few seconds, if the forecast stays on the ground while the phone moves, is calibrated, and needs only a monocular camera and visual-inertial odometry (VIO). We build and evaluate such a system. People are detected, lifted onto the floor by ray-plane intersection, tracked in a gravity-aligned metric frame, and forecast as per-step probability maps by a small U-Net trained with a negative log-likelihood (NLL) loss on bird's-eye (SDD) and first-person (EgoTraj-Bench) trajectories. On handheld ADVIO recordings, vertical VIO drift and the user's own changes of level silently rescale monocular ground positions (by 87% within 90 s on one clip; on another, all tracks are lost for the last 31% of the clip); keeping the camera's height above the floor constant under a low-pass-filtered altitude avoids this, though it lags on escalators. On the SDD and EgoTraj-Bench test splits, the final forecaster lowers the NLL at 4.8 s by 1.51 and 1.37 nats relative to a fitted constant-velocity Gaussian. Lacking ground truth for people in handheld video, we score forecasts against the tracker's own later raw measurements. In an internally pre-registered evaluation on seven held-out clips, the final forecaster's NLL is lower than the benchmark-fitted baseline's at 1.2, 2.4 and 4.8 s (by 0.17, 0.23 and 0.47 nats; 95% intervals over people exclude zero), but by less than half as much as on the development clips. Exploratory analyses cut both ways: resampling clips instead of people widens the intervals to include zero at 1.2 and 2.4 s, and once both forecasters are recalibrated on the development clips the network is significantly better only at 1.2 s; but two clips run with ADVIO's reference poses favour the network much more when re-run with the phone's own poses. We discuss what such self-consistency scores can and cannot show.

Concerns are growing about the water use of data centers. To quantify the scales, drivers, and possible trajectories of this water use, analytical estimation is required. However, current estimation methods and results vary widely due, in part, to a lack of established best practices. This variance is illustrated with a U.S. case study demonstrating a roughly 24-fold difference in possible outcomes based on observed U.S. study design variations. To address these challenges, this review synthesizes 14 best practices for advancing the science and utility of data center water use estimation. These proposed best practices can also be used as critical questions for stakeholders when evaluating the quality of any study. Using a structured rubric, it further assesses the literature to date (31 studies) against these best practices. The assessment reveals important opportunities for improvement and standardization in future study scopes, clarity, methods, and scientific knowledge building utility. It further finds that, while many studies provide reasonable transparency, scope definitions, replicability, and uncertainty treatment, many fall short in scientific justification, data representativeness, transparency on hydropower assumptions, terminological consistency, and meaningful discussions of limitations and future work. Recommendations are proposed for analysts, stakeholders, and policymakers to leverage these findings to help build a vibrant data center water use research community. Finally, this review can also be used as a basic primer for understanding data center cooling configurations and water use among analysts, policymakers, the media, and the public.

Quantum communication can outperform classical communication in a variety of information-processing tasks, but such advantages are typically established for a fixed communication network. Here we introduce topology-independent quantum advantage, in which a quantum protocol outperforms every classical protocol subject to the same communication constraint, irrespective of how the classical communication resource is distributed among the parties or which communication topology is employed, even when the parties may share unlimited classical randomness. We first demonstrate this phenomenon through a simple three-party equality task: two senders receive ternary classical inputs, while an input-free receiver determines whether the inputs are equal. We then formulate a general framework for topology-independent communication and completely characterize the corresponding classical correlation polytope in the minimal three-party scenario. Its facet inequalities reveal several instances of topology-independent quantum advantage. Remarkably, one of these inequalities enables the semi-device-independent certification of a nonunitary quantum channel, while another certifies a sequential quantum network topology without any prior assumption about the network configuration. Finally, we extend the framework to four-party networks, characterize the relevant classical communication structures, and identify a communication task exhibiting quantum advantages against all admissible classical topologies. Our results establish topology independence as a genuine and robust form of quantum advantage in communication networks.

Mean-field control barrier functions (MF-CBFs) enforce safety of large-scale multi-agent systems at the level of densities, but existing formulations require the control to act directly on the population to be kept safe. We consider populations that are not directly actuated and are influenced only through other, controlled populations, possibly in the presence of diffusion. We show that if the constrained population lies at directed graph distance d from the actuated one, then, under a nondegeneracy condition, the safety functional has relative degree d+1 and the control enters its (d+1)-th time derivative affinely. Higher-order safety constraints can therefore still be enforced through a convex quadratic program. We apply the framework to leader-follower density control among dangerous regions, where the follower safety functional has relative degree two and the leader safety functional has relative degree one. We characterize exactly when the two constraints are jointly infeasible and bound the tracking errors induced by the safety correction. Simulations in cluttered environments show that the proposed filter preserves follower safety, whereas enforcing leader safety alone does not.

This paper develops a one-dimensional model for a mobile viral component propagating through a spatially distributed susceptible population. Random viral motion produces diffusion, whereas a response to the relative gradient of susceptible density generates directed transport. The existence of a positive entire traveling wave is proved when directed transport dominates diffusion, and the limiting case of equal transport and diffusion coefficients is constructed separately. A dimensionless transformation then converts the remaining equation into a logistic equation and supplies closed-form profiles. These formulas identify the unique viral maximum and the far-field behavior. A conservative implicit-explicit finite-difference method provides an independent numerical check. Diffusion is evaluated at the new time level, the gradient-driven flux explicitly at cell interfaces, and susceptible depletion pointwise. Exact traveling-wave values specify the initial data and time-dependent Dirichlet conditions at both endpoints. Thus, the computation tests preservation and translation of the analytical profiles. A calculation with eight hundred spatial subintervals reproduces the localized viral band and the monotone susceptible front. The largest uniform errors are approximately $1.70\cdot10^{-2}$ for viral density and $2.22\cdot10^{-3}$ for susceptible density. The numerical results corroborate the analytical construction and show that the conservative discretization captures the proposed propagation mechanism.

The SU$(N)$ matrix-model potential $V=-\tfrac{g^2}{4}\sum_{I,J}Tr[X_I,X_J]^2$, the interaction of the BFSS/BMN family, whose thermal large-$N$ states are dual to black holes, is diagonal in the coordinate basis and is compiled in the literature as a phase polynomial with $O(d^2N^4Q^4)$ rotations per Trotter step. That cost is close to optimal for its gate class: we show that for a generic Trotter step the Walsh support $m_{\min}$ lower-bounds both the rotation count and the two-qubit count of every circuit over $\{CNOT,CZ,CPhase,R_{ZZ},SWAP,X,R_z\}$, ancillas included, and we build a parity network attaining $1.01$–$1.10\,m_{\min}$. Improvement therefore requires a gate outside that class, and the exact factorisation of $V$ into a sum of squared commutator entries supplies one in each era. For NISQ, Hadamard: each squared entry is added into a Fourier register and squared there, $\Theta(d^2N^3Q^3)$ two-qubit gates. For fault tolerance, $T$: reversible arithmetic with one phase-gradient injection per squared entry, $\Theta(d^2N^3Q^2)$ T gates and no arbitrary-angle rotation. Both are built and verified. With a native $R_{ZZ}$ the NISQ circuit uses fewer two-qubit gates than every parity network from $Q=6$ at SU(2) and $Q=2$ at SU(5); the oracle beats a precision-matched Hamming-weight-phasing incumbent at every point built except the smallest, by up to $11.9\times$ in T count and, at equal width, $25.5\times$ in surface-code spacetime. An ablation with the same arithmetic but no factorisation costs up to $21\times$ more and loses to that incumbent at $15$ of $17$ points, so the factorisation, not the arithmetic, produces the win. Both constructions survive the supersymmetric deformation to mini-BMN with fermions, and complete Trotter evolutions on a statevector simulator reproduce the exact product formula to $2\times10^{-12}$.

We investigate the explicit geometric and combinatorial structure of Mirković-Vilonen (MV) cycles in twisted affine Grassmannians associated with special parahoric group schemes. Using the twist map on configuration spaces of decorated flags, we obtain an explicit parametrization of twisted MV cycles by Lusztig data. We show that the changes between reduced words are governed by the usual Lusztig transformations for the reductive group arising from ramified geometric Satake correspondence. As a consequence, the moment polytopes of twisted MV cycles coincide with the usual MV polytopes for this group. Our construction applies to all special parahoric subgroups in the tamely ramified setting.

In this work we study the Bell-CHSH inequality in the vacuum of the free massive Majorana field in $1+1$ dimensions, taking as the dichotomic observables of the two parties even bilinears of smeared fields localized in the right and in the left Rindler wedges. The test functions are generated from seeds in the modular-frequency variable through the Tomita-Takesaki operator of the right wedge and, on the side of the left wedge, through its adjoint multiplied by $i$, which accounts for twisted duality. For seeds supported at positive modular frequency the one-particle overlaps between the two parties do not depend on the modular weights, and we obtain the Bell parameter in closed form. When the seeds are phase rotations of a single compact bump, the bilinears generate a copy of the Pauli algebra in each wedge, and we compute exactly the correlation matrix of the vacuum restricted to the two copies. The optimal settings violate the classical bound up to spectral widths more than twice as large as those reached with planar settings. We also show that the Tsirelson bound is never attained in the wedge vacuum for any choice of dichotomic observables, and that approaching it forces the modular spectral weight of the Bell vectors to concentrate at the spectral point $\lambda=1$ of the modular operator. This establishes in a precise way how the even sector of the free Majorana field realizes the Summers-Werner mechanism.

Ultra Reliable Low-Latency Communication (URLLC) demands submillisecond latency and reliability above 99.999%, which are difficult to guarantee under dynamic wireless and atmospheric conditions. Although free space optical (FSO) links provide high-capacity transmission, their performance degrades under turbulence and fog, while radiofrequency (RF) links offer robustness but limited bandwidth. This paper presents a reinforcement learning-based hybrid optical RF framework that adaptively selects transmission modes to satisfy stringent URLLC requirements. The hybrid decision process is modeled as a Markov Decision Process and optimized using a Deep Q Network (DQN) controller. Extensive simulations under varying attenuation and fading scenarios demonstrate up to 35% latency reduction and 25 to 30 % throughput improvement compared to RF-only and static hybrid schemes, while maintaining reliability above 99.999%. A laboratory-scale testbed operating at 1550 nm and 28 GHz further validates the approach, showing less than 6% deviation between simulated and measured results. The proposed framework demonstrates the practical feasibility of intelligent hybrid transmission for mission-critical next generation wireless networks.

In worker-equipment proximity monitoring, video transformers are widely used for collision anticipation and have demonstrated strong performance. However, their accuracy comes with substantial computational demands, creating a tension with the need for low-latency inference on mobile robots and the pursuit of lower-carbon computation in construction. To address this, this study investigates where computation within an established video transformer is redundant and whether that redundancy can be removed without materially degrading predictive performance. Using VideoMAEv2-Base on the Nexar Collision Prediction dataset, we first examine how collision-relevant information evolves across network depth and then investigate two complementary forms of redundancy: structured capacity redundancy in multilayer perceptrons (MLPs) and spatiotemporal redundancy in the token stream. Linear probes show that interpretable motion cues, including flow magnitude, looming, and approach versus retreat, are most accessible at intermediate layers, whereas collision-label discrimination strengthens toward the final layer. Token redundancy is axis-specific: adjacent temporal-token similarity reaches 0.970 in later layers, while spatial similarity falls to 0.297, indicating substantially greater redundancy across time than across space. Exploiting this asymmetry, temporal token merging reduces backbone computation from 356.99 to 178.50 GFLOPs and latency from 12.69 to 7.21 ms per clip, a 1.76x speedup, while mean average precision changes only from 0.7478 to 0.7443. Importance-guided retention of 50% of MLP units preserves an AUC of 0.753, compared with 0.529 under matched random retention, and reveals that pruning alters score calibration before discriminative ranking collapses. These findings establish a new pathway for pursuing faster algorithms through targeted temporal token compression and neuron pruning.

Full actuation is commonly valued for independent position and attitude tracking. This work establishes a complementary principle: when only a position trajectory is prescribed, minimizing a positively homogeneous allocation cost selects underactuated-like behavior on a fully actuated multirotor. This cost class includes norms and power laws, including standard minimum-norm pseudoinverse allocation. We use fiberwise infimization to transport the actuator-output cost to wrench space, yielding a homogeneous infimal allocation cost. Restriction to unit zero-moment forces defines a directional landscape whose minimizing directions are independent of force magnitude. We show that aligning any fixed minimizing direction with the required world-frame force minimizes instantaneous and integrated zero-moment allocation costs. The resulting body force remains in a fixed one-dimensional subspace without transverse components. We derive an exact optimality-gap decomposition quantifying the additional cost of every other attitude trajectory and characterizing the minimum-cost class. When a disturbance or prescribed-motion change modifies the required-force direction, a fully actuated platform can immediately generate the modified force using transverse authority and reorient toward the new minimum-cost family. A structurally underactuated platform must reorient before generating that force. Thus, homogeneous allocation-cost minimization explains why optimal nominal motion can be underactuated-like, while full actuation remains valuable for disturbance rejection and transient force generation.

Structured prediction tasks pose unique challenges for in-context learning (ICL): their compositional outputs require modeling fine-grained, token-level patterns that sentence-level approaches fail to capture, and their task-specific annotation conventions are human-defined artifacts that cannot be acquired through pretraining alone. We propose Verb-ICL, a selective annotation framework for ICL-based structured prediction that addresses both challenges. Verb-ICL first selects representative examples using a token-level coverage strategy that captures local semantic patterns critical for structured prediction, then generates actionable error feedback that codifies task-specific annotation guidelines and incorporates this feedback into ICL demonstrations. We evaluate Verb-ICL on six structured prediction datasets spanning information extraction and semantic parsing. Experiments with recent LLMs show that Verb-ICL consistently outperforms strong selective annotation baselines under low-resource settings and continues to provide gains as the annotation budget increases. Extended analyses demonstrate that the generated feedback is predominantly useful across a four-category quality taxonomy, generalizes as task-level guidance beyond instance-specific corrections, and improves performance regardless of the underlying selection strategy.

Two physical consequences of the recent topological classification of 2d Fermi seas for free particles based on flat orbifolds $\mathbb{R}^2/\Gamma$ of discrete symmetry $\Gamma$, are explored. The first is a study of the spatial density behavior inside the mesoscopic window enclosed by the Fermi and (infinite) thermal lengths. It is shown that the single-particle density matrix exhibits a sharp directional anisotropy, encoding a real space diffraction pattern, i.e., a Fraunhofer-like diagram of the Fermi sea, that reveals the point-group symmetries of the background lattice $\Gamma$ in the form of anisotropic Friedel oscillations. As examples, the special cases of the $p4m$ and $p6m$ crystallographic flat orbifolds are treated in some detail. The second physical consequence is the establishment of an explicit geometric correspondence between momentum-space packing and electrical transport. A mean-field expression for the longitudinal electrical conductivity $\sigma_{xx}$ as a function of the isoperimetric parameter $\cal{I}$ of the occupied orbifold domain by utilizing geometric measure theory is presented, satisfiying the scaling law $ \sigma_{xx} \propto 1/\sqrt{\cal{I}}$. The topological transport formula naturally incorporates the vanishing conductivity ($\sigma_{xx} \to 0$) of the previously described “geometric insulators,” promoting the isoperimetric parameter as an efficient descriptor for topological and geometric phases in conductors.

This work proposes a latent Lagrangian-based framework for reduced-order modelling of forced nonlinear dynamical systems. In contrast with conventional Lagrangian or Hamiltonian neural networks, our approach learns a set of latent coordinates sufficient to capture the dynamics conjointly with two neural networks for the latent kinetic and latent potential energies, and leverages force supervision to eliminate the need for an ODE solver during training. Consistency of physical laws in the latent space is ensured through the principle of virtual work. Results show that the model effectively learns the subtle dynamics induced by the system's nonlinearity and non-convex potential energy, and generalizes well to unseen forces and initial conditions. These observations confirm the physical relevance of the proposed approach, and its interest for model reduction.

Sparse-view novel view synthesis is a central problem in 3D content creation, but diffusion-based approaches remain limited by iterative denoising, making multi-view generation expensive at inference time. We introduce NAMVIS, a diffusion-free framework that reformulates multi-view image synthesis as geometry-conditioned next-scale autoregression. Instead of generating target views through repeated denoising, NAMVIS predicts discrete visual tokens through a small number of coarse-to-fine scale steps, while sampling all tokens within each scale and across target views in parallel. To anchor this generation process to explicit camera geometry, we propose Multi-scale Projective Pose Encoding, which injects source and target camera transformations into both target-view self-attention and source-to-target cross-attention at every resolution. NAMVIS further combines global conditioning with dense geometry-aware cross-attention, enabling the model to preserve source-view appearance while maintaining target-view consistency. Across Objaverse, GSO, and OmniObject3D, NAMVIS outperforms diffusion-based baselines in PSNR, SSIM, and LPIPS, while running over 3 times faster than the evaluated diffusion baselines under the same evaluation setting. These results suggest that geometry-conditioned next-scale autoregression is a promising and efficient alternative to diffusion for sparse-view multi-view synthesis. Additional qualitative results, videos, and resources are available at https://corl-team.github.io/namvis/

Structural diagrams are widely used to represent complex systems and relational information across scientific, engineering, procedural, and spatial domains. Recent vision-language models (VLMs) have become increasingly capable of recognizing diagram elements and reasoning about their content, while complete diagram topology extraction remains comparatively underexplored. In this paper, we study diagram-to-graph topology extraction: extracting all diagram entities and the complete relations among them. To enable large-scale supervised training and systematic evaluation of this task, we introduce Knossos, a benchmark of 19,200 diagrams across six diverse domains, with 245,179 nodes and 439,740 edges. Its symbolic generation process provides exact alignment between rendered diagrams and annotations of complete topology, relation types, and connector geometry. To address the modeling challenge of complete topology extraction, we also present Ariadne, a structured framework that decomposes the task into node inventory extraction and source-conditioned edge prediction. Extensive experiments show that training on Knossos substantially improves complete topology extraction in smaller open-source VLMs. Ariadne further improves over one-step extraction under matched supervision, demonstrating the additional benefit of structured decomposition. It achieves the highest average Edge F1 among the evaluated methods on Knossos, while both backbone variants also improve over their unadapted counterparts on the real-world external benchmark. Code and benchmark are available at https://github.com/bangwayne/knossos_Ariadne_Public.

In a Transformer, token mixing is the step that lets each token draw information from other tokens, and it dominates the cost of encoding long sequences. Self-attention does this mixing very well: every token weighs every other token by content, which gives strong contextual modeling. That all-pairs comparison is also why its cost grows quadratically with sequence length. We introduce WNet, a Transformer encoder that replaces self-attention with token mixing based on the discrete wavelet transform (DWT). Three attention-free mixers recombine the scales: by linear fusion, by learned gating, or by letting each token choose its scales. A hybrid adds self-attention in the last layer only. A receptive-field analysis shows that wavelet mixers built from two-tap filters, such as Haar, never relate tokens outside fixed blocks, however deep the network, even when the filters are learned. Longer filters reach the whole sequence within two layers. We pre-train every model with masked language modeling on a fixed-token subset of C4 and fine-tune on GLUE, using one controlled setup with size-matched BERT and FNet baselines and a control that cannot mix tokens. The token-gated mixer trains as fast as attention at 256 tokens and 2.7 times faster at 4,096.

AI assistants receive requests that leave out information needed for a good outcome, for example about users' preferences or goals. They must then either speculate or ask for more information before proceeding. We reconceptualize this as a value-of-information problem: the assistant should acquire information whose absence causes the greatest avoidable loss in user utility. This is rarely known ex ante; rather, assistants must predict it in order to optimally allocate limited user interactions. We instantiate this problem in image generation and derive a reinforcement learning framework using multi-turn simulated users to maximize utility recovery under uncertainty. In a preregistered study with 456 interactive sessions across 76 human participants, this helped users significantly better match reference images with significantly fewer questions, less total interaction time, and lower cost. This points toward a simple and scalable framework for training language model assistants to better disambiguate user intent by asking more informative questions.

Most federated learning (FL) studies implicitly assume that training data is readily available at participating edge devices. In practice, this data must first be acquired through heterogeneous sensors with potentially different measurement modalities, and the quality and cost of this acquisition process can directly affect the learning performance. This dependence creates a coupled relationship among sensing, computation, and communication, necessitating a system model that captures their joint effect on resource efficiency and learning accuracy. This paper develops such a unified treatment for an over-the-air (OTA) FL system in which each device configures its sensing modality, power, resolution, and sample size, with aggregations conducted OTA at the server. We derive models for sensing noise, energy consumption, latency, and OTA aggregation distortion, and establish a non-convex convergence bound quantifying how these factors influence the global model trajectory. Guided by this analysis, we formulate a per-round optimization that balances energy, latency, and learning quality under physical device constraints. The resulting mixed optimization is addressed through an alternating procedure with closed-form and one-dimensional updates for continuous variables and a lightweight discrete search. Numerical experiments corroborate the advantage of jointly optimizing sensing, computation, and communication decisions.

In this work, we investigated the absorption of light in perovskite nanowires (PrvNWs) for application as active layer in a solar cell. Specifically, we simulated PrvNWs with a random position inside the pores of an AAO membrane. The results demonstrate that this vertical disorder can boost light absorption, up to 28% in terms of short-circuit current, in comparison to the corresponding ordered PrvNW array. This performance boost is due to a combination of light trapping and anti-reflection effect. Thus, our findings suggest that vertically disordered PrvNWs can have a much better performance than ordered arrays. Also, we showed that it is not necessary to aim for a very narrow length distribution in the nanowires, easing the fabrication requirements.

Logging code supports debugging, monitoring, and software maintenance, but excessive logging can add noise, impose runtime overhead, and obscure diagnostic information. While prior research has extensively studied logging code generation and modification, logging removal remains comparatively underexplored. In this paper, we study developer logging removal practices and explore the use of coding agents for this task. We extract and manually validate logging removal cases from Python and Java repositories and derive 10 removal patterns and 11 removal reasons that characterize how and why logging code was removed in real-world software changes. We further construct LogRem, a dataset of 387 real-world cases covering direct logging statement removal, logging infrastructure removal, and logging replacement. We evaluate four coding agents with multiple model settings and compare their outputs with accepted real-world changes. Although 95.6% to 100.0% of outputs pass validity checks, only 11.1% to 19.6% remove the same logging code as the corresponding real-world change while preserving unrelated code. Agents differ through missed removals, extra removals, and unrelated code edits, with substantial variation across logging removal categories and trajectories. Execution cost varies widely, but higher cost does not consistently yield closer alignment. Commit messages and developer discussions provide the largest alignment gains, while taxonomy guidance consistently reduces runtime. Overall, our study establishes logging removal as a distinct software maintenance task and shows that reliable automation depends on accurately determining removal scope while preserving necessary code. To the best of our knowledge, this is the first study to examine logging code removal from this perspective.

We present GPU-accelerated Bregman Douglas–Rachford splitting algorithm (BDRS) for discrete optimal transport problem in three input formats: an explicit cost matrix, a point cloud with a ground cost between them, and a separable cost on a regular grid. For each input format, we propose hardware-aware designs of mathematically equivalent representations for the BDRS iterations to enhance numerical stability and empirical runtime. We benchmark the three proposed implementations against eight GPU baseline solvers from the literature on the same device. We demonstrate that our implementations of BDRS achieve state-of-the-art performance on their respective input formats. To the best of our knowledge, this is the first cross-solver study of GPU DOT solvers with a unified measure of optimality.