Papers

Sycophancy in large language models risks amplifying unsupported claims or beliefs across the knowledge ecosystem as these systems increasingly contribute to research synthesis, evaluation, and decision-making. A reliable model should revise its judgement when a user identifies a genuine flaw, but should resist pushback based on fallacies, emotion, or irrelevant appeals to authority. While sycophancy is usually measured on questions with a verifiable answer, the evaluative tasks in which it is most difficult to detect have no such ground truth. We introduce Big-Big-SycoBench, which measures sycophancy through argument discernment: the ability to revise an evaluation after valid criticism while resisting invalid pressure. The benchmark consists of 450 text artefacts with 5,400 arguments for raising or lowering their quality scores, each reviewed by one or more of twelve human annotators. An LLM scores each artefact twenty times to establish its baseline score and uncertainty. The model is then given a valid or invalid argument and asked to update its score. We only count a revision if it exceeds the model's own baseline variability. We use the resulting score delta to distinguish responsiveness to valid arguments from compliance with invalid ones. Across ten models, responsiveness is 88-95% for seven, 76-77% for Kimi K3 and GLM-5.3, and 32% for Opus 5.5. Compliance ranges from 0.2% for GPT-6 Astra to 95% for GPT-4o. Big-Big-SycoBench is intended to guide the development of models whose judgements can be corrected by sound arguments, but not displaced by unsupported pressure.

Propositional Dynamic Logic (PDL) is a well-established modal logic of programs. Among its many extensions, the loop operator stands out for capturing the cyclic behaviour of programs: a world satisfies loop(pi) precisely when it can return to itself along a path matching the regular expression pi. While the satisfiability problem for the resulting logic LPDL was settled already in the 1980s, the decidability of its finite satisfiability problem has remained open ever since. The core difficulty is inherent to the finite setting: requiring and forbidding cycles at once is easily reconciled over arbitrary models, but over finite ones it defeats the existing finite-model-theoretic techniques. We make the first substantial progress towards closing this gap. We prove that finite satisfiability is ExpTime-complete for LPDL formulae (also in the presence of the converse operator) in negation normal form, under the restriction that every negated non-atomic loop has the form loop(pi^+) for some arbitrarily complex program pi – a generalised form of self-reachability; positive occurrences of loops remain unrestricted. At the heart of our approach lies a novel way of representing intended finite models as their infinite, cactus-like unravellings: these are recognisable by two-way tree automata and satisfy a tailored omega-regular foldability condition that lets us finitise the accepted structures.

Codes in the Levenshtein metric that can correct multiple edits (including deletions, insertions, and substitutions) have attracted a lot of attention recently owing to their application in DNA-based storage and racetrack memories. In this work, we present new techniques for designing explicit binary codes of length $n$ that correct $t$ edits with at most $(2t+1)\log n+O_t(\log\log n)$ bits of redundancy. We first propose to use a polynomial hash to construct a systematic code that corrects $t$ edits with $(4t+2)\log n+O_t(\log\log n)$ bits of redundancy. Then we observe that there are many polynomial hashes like this, some are good, some are bad. So, we propose the second technique, called entropy recycling, to use the indices of good hashes to carry information. As a consequence, we obtain a partial-systematic code of length $n$ that corrects $t$ edits with at most $(2t+1)\log n+O_t(\log\log n)$ bits of redundancy. Finally, we provide deterministic polynomial-time encoding and decoding algorithms for our code.

Choosing a mixed-precision configuration for large language model (LLM) serving trades speed against accuracy. Allocators that use average bit-width or a memory budget as a speed proxy can miss changes in the resource that limits execution. We argue that served speed can be limited by compute, weight loading, or key-value (KV) cache access, depending on the phase, batch size, and context length. A byte-based proxy does not capture all of these costs, so fewer bits do not always mean faster serving. To address this gap, we build a cost model that accounts for the binding resource and use it in two ways. First, as a predictor: given a configuration and operating point, it estimates serving latency using constants calibrated for the model, GPU, and serving stack. We evaluate its accuracy across three hardware classes. Second, as an allocator: we combine modeled time costs with measured per-tensor accuracy costs and use a dynamic program to select a configuration under a latency or accuracy budget. We compare its configurations against byte-aware and sensitivity-only allocation.

Deep learning is now central to radiology, helping detect changes on X-rays, CTs, and MRIs. However, these systems are highly vulnerable to adversarial attacks - small, crafted perturbations that mislead models while appearing unchanged to humans. Such errors risk missed cancers or false positives. Studies show over 90% attack success on medical scans. Almost 30% of US hospitals use AI in imaging in 2023. Thus, an effective method of detecting such attacks and their directionality is crucial to ensure the credibility and reliability of medical systems based on DNNs. We propose a novel detection framework that leverages subsequent attacks using random noise to identify adversarial perturbations. Our approach uses the analysis of variations in the Sigmoid function to distinguish genuine medical images from adversarially manipulated ones. Specifically, we compare prediction differences between clean and attacked images, as well as between different adversarially altered versions, to improve detection accuracy. We evaluated our method on three popular medical datasets including 'Chest X-Ray Images for Classification', 'Retinal Fundus Multi-disease Image Dataset' and 'Brain Tumor Dataset' under various attack scenarios, including random noise. Our framework achieved up to 99.8% ACC in identifying adversarial directional attacks, significantly outperforming existing defense mechanisms. It consistently identified adversarial samples across varying attack strengths while keeping false positives low, showing strong reliability and potential for clinical use. Furthermore, our proposal highlights its potential for real-world deployment.

Electroencephalography (EEG) recorded for motor-imagery (MI) brain-computer interfaces (BCIs) carries subject-identifying structure that persists across sessions, allowing separately released sessions from the same person to be linked. Existing EEG identity protections use learned or additive perturbations and operate on representations other than the covariance geometry underlying MI pipelines. We ask whether cross-session identity leakage can be suppressed directly in this covariance representation without learning a privacy model. We introduce a Riemannian transform that equalizes two subject-conditioned geometric statistics: the Riemannian mean covariance and the tangent-space scatter around it, with two optional band-specific refinements. Ablations show that mean alignment and scatter whitening carry nearly all of the suppression, while the optional stages contribute small, dataset-dependent refinements. Across two public datasets and five attackers spanning four covariance-space representations, the worst attacker falls from 90.8% to 15.6% on BCI Competition IV-2a and from 90.8% to 9.9% on OpenBMI, at a cost of 7.0 and 2.1 points of MI decoding accuracy relative to the 8-30 Hz input, and no aggregation rule exceeds chance by more than two points at any batch size. The transform requires no learned privacy model, surrogate network, or gradient-based training, and is fitted on one session and applied unchanged to a held-out later session.

World Action Models (WAMs) unify visual world modeling and action generation, yet their real-world deployment is heavily bottlenecked by iterative action denoising and the sequential execution of visual prefilling and action prediction. We demonstrate that the action-denoising trajectories in WAMs are approximately straight, rendering the first-step velocity a strong estimate of the final action direction. Building on this observation, we introduce 1Step-WAM, a framework that transforms multi-step action denoising into one single forward. Specifically, we employ a lightweight calibrator to directly correct the first step velocity, and use bandwidth-constrained action decoding to further reduce errors in the calibrated one-step actions. Compressing action generation into a single step enables it to execute in parallel with visual prefilling, completely breaking the sequential bottleneck of the traditional imagination-then-action paradigm. Evaluations on RoboTwin and LIBERO show that 1Step-WAM largely preserves multi-step task success with only about 13.4M trainable parameters, while achieving inference speedups of $7.44\times$, $5.98\times$, and $24.84\times$ for FastWAM, ImageWAM and Kairos, respectively. Furthermore, real-world robotic deployments using the FastWAM backbone confirm a consistent 7.44$\times$ speedup while maintaining close success rate to the multi-step baseline. Code is avaliable at https://github.com/changdong1687/1Step-WAM.

Hybrid federated semantic communication enables user equipment (UEs) to contribute model updates or semantic features under a shared training framework. However, heterogeneous role costs and the interaction between model aggregation and feature reconstruction (FR) remain key challenges in practical radio access network (RAN) deployment. To address these issues, we propose ARC-FedSC, a RAN-aware framework that coordinates UE role assignment and server-side controlled FR under a fixed communication budget. Within ARC-FedSC, RAN Aware Role-Bandit learns the historical contribution of UEs in different roles together with current uplink conditions, while Temporal Consistency Alignment (TCA) regulates FR according to recent model aggregation directions. Experiments on CIFAR-10 with time-varying RAN channels show that ARC-FedSC outperforms baselines across all five tested role configurations in the later training stage, with the best configurations achieving about 24.17 dB average PSNR over the last 20 rounds. Component level results further show that Role-Bandit maintains an average rolling UE coverage of 85.3%, compared with about 65.7% for baseline, while TCA reduces the probability of harmful FR events from 0.245 to 0.093, corresponding to a 62.0% reduction. Ablation and sensitivity studies further support the effectiveness and robustness of the proposed design.

Multimodal Chain-of-Thought (MCoT) substantially enhances the complex reasoning capabilities of Multimodal Large Language Models (MLLMs) by explicitly modeling intermediate reasoning processes. However, most learning-based MCoT methods follow a stage-centric paradigm, where different reasoning behaviors are assigned to predefined stages and optimized separately. This results in fragmented reasoning and a training-inference reasoning-state gap. To address these limitations, we propose ReState, a unified iterative reasoning framework that reformulates multimodal reasoning from a stage-driven pipeline into a state-driven dynamic process. It comprises two complementary components: Unified Reasoning State Modeling (URSM) and Closed-Loop Reasoning-State Learning (CRSL). URSM unifies reasoning construction, correct reasoning preservation, and erroneous reasoning correction as reasoning-state transitions within a shared state space, enabling these heterogeneous reasoning behaviors to be learned jointly. CRSL further incorporates self-generated historical reasoning states during training and recursively updates the current state during inference, thereby mitigating the reasoning-state gap and enabling iterative self-refinement without requiring an additional reflection module, critic, or independent correction stage. Furthermore, we construct VersaCoT-10K, a high-quality CoT dataset spanning eight visual reasoning domains, which provides diverse supervision for multi-domain reasoning and reasoning-state refinement. Extensive experiments demonstrate that ReState exhibits strong adaptability across different MLLM backbones and achieves strong performance on multiple public reasoning benchmarks, supporting the shift from stage-centric reasoning toward state-centric reasoning. Our data and code are available at https://github.com/YuChuang1205/ReState

Recursive synthetic-data training can degrade a fitted model, but its effects on the variable-screening stage that precedes model fitting are less understood. When each generation labels data for the next, a marginal screener ranks variables against the current pseudo-label target rather than the original response. Across generations, this target drift can erode or reverse the active–inactive score margin and cause screening to discard variables that clean-label screening would retain. We call this process screening collapse. In a Gaussian linear model with minimum-norm recursive generators, we establish finite-sample screening bounds, characterize a late-generation information-loss limit, and derive a minimax impossibility result. Screening collapse can precede and compound ordinary model collapse: omitted variables incur a covariance-weighted loss that no refit restricted to the selected variables can eliminate, while recursive labels further degrade estimation on the retained variables. We also examine two approaches to mitigating screening collapse. Fixed-fraction real-label retention preserves a clean-label component but leaves resolution fixed, whereas data accumulation can avoid screening collapse if the pooled target preserves a positive margin and the growing sample resolves it. Numerical experiments on gene-expression data support the theoretical findings.

Dietary transitions away from animal-based meat could reduce food-related environmental burdens, but plant-based alternatives must compete on flavor, texture, and overall liking. Here we extend a generative AI framework with 469 plant-based burger recipes and use it to create six formulations across five primary ingredients: bean, chickpea, lentil, mushroom, and oat. An executive chef translates the AI-generated formulations into standardized preparation protocols and refines the ingredient quantities based on culinary expertise. The integration of generative modeling and culinary expertise defines a human-centered generative AI framework for food design. We evaluate the six burgers in a blinded sensory survey with 101 participants in an operating restaurant and quantify their land use, greenhouse gas emissions, eutrophication potential, and scarcity-weighted water use. We use Big Mac scores from a previous survey as an iconic commercial benchmark. The mushroom burger achieves mean flavor, texture, and overall liking scores similar to the Big Mac, with less than one-third of its estimated environmental impact. The other formulations reduce the estimated impact further, in several cases to near one-tenth of the Big Mac value, but receive lower liking scores. Mixed-effects models quantify associations between sensory attributes and human liking and retain predictive accuracy on independent survey data, with mean absolute errors no greater than 0.52 points on a seven-point scale. Together, these results demonstrate how generative design, culinary expertise, human sensory response, and environmental assessment can converge to create sustainable foods that people want to eat.

Multi-label learning typically pursues model construction through the joint optimization of label fitting, correlation modeling, and structural regularization. While empirically effective, such a coupled paradigm imposes multiple learning objectives on a single predictive parameter matrix, forcing it to simultaneously fit observed labels and encode structural priors. This coupling may induce ill-conditioned optimization geometry, increase hyper-parameter sensitivity, and complicate capacity control. To address the these issues, this paper proposes a Progressive Multi-Label Learning (ProML) framework, which decouples direct label fitting from predictor learning via the introduction of an auxiliary predictive variable. This auxiliary variable admits a closed-form solution, yielding an equivalent profiled objective in which the supervised fitting term is scaled by a trade-off parameter. This result provides a unified view of ProML and conventional coupled learning, with the latter recovered as a limiting case. We theoretically analyze the optimization geometry of ProML and show that, under quadratic positive semidefinite structural regularization, ProML attenuates the dominant curvature of the fitting term, enlarges the stable learning rate range, and provides a controllable fidelity-regularization trade-off. Further, we instantiate the ProML for multi-label feature selection task by integrating global-local label priors and row-sparse regularization. Extensive experiments and controlled optimization analyses on 12 benchmark datasets empirically corroborate the theoretical analysis and demonstrate the robust and competitive performance of the proposed method.

Effective unsupervised domain adaptation (UDA) requires a representation that captures transferable structure from both the labeled source domain and the unlabeled target domain. We introduce Joint Contrastive-Adversarial Representation Learning (JCAR), a two-stage framework that learns a common representation from the union of unlabeled source and target samples by combining contrastive learning with adversarial domain alignment, and then freezes the representation to fit a source-supervised predictor. We establish a non-asymptotic excess-risk bound for this procedure that couples representation and domain-discriminator estimation while separating downstream prediction, with explicit neural-network approximation errors. Under the stated assumptions and common Hölder smoothness $\beta\geq1$, the bound identifies a sufficiently large unlabeled-sample regime in which the downstream rate is $\widetilde{\mathcal{O}}(m^{-\beta/(2\beta+d)})$, where $m$ counts labeled source samples and logarithmic factors are suppressed. The labeled-sample exponent is governed by the latent dimension $d$ rather than the ambient dimension $p$, which enters the pretraining cost. Experiments on Office-31, Office-Home, and DomainNet demonstrate clear transfer gains across diverse domain shifts, with particularly strong performance when representations are learned from scratch.

Accurate lesion segmentation in mammography is important for computer-aided breast cancer diagnosis, but dense pixel-level annotations remain costly and limited. Existing public segmentation datasets usually cover only a small set of lesion types with manually delineated masks, whereas larger mammography datasets often provide only image-level diagnostic labels. This annotation mismatch makes it difficult to train segmentation models that can generalize to rare or weakly annotated lesion categories. To address this problem, we study open-vocabulary mammographic lesion segmentation under a mixed-supervision setting, where pixel-level masks are available for base lesion categories and image-level labels are available for a broader set of categories. We propose a two-branch framework that jointly learns from segmentation data and classification data. The segmentation branch learns lesion boundaries from annotated masks, while the classification branch introduces additional image-level semantic supervision. To better transfer image-level supervision to lesion-relevant regions, we further introduce a Lesion-wise Attentive Classifier that maintains category-specific feature prototypes and generates lesion-aware attention maps for foreground feature pooling. This design reduces the influence of background tissue and encourages more compact lesion representations. Experiments on multiple mammography datasets show that the proposed method consistently improves open-vocabulary segmentation baselines on both base and novel lesion categories, while reducing the gap to fully supervised models that require dense target-domain masks. Qualitative results further indicate that our method produces more complete lesion localization for small, low-contrast, and visually subtle mammographic findings.

We consider the recovery of functions of the weighted Wiener classes $W_s^\mu$ from perturbed Fourier–Hermite coefficients in the metric of $L_\infty(\mathbb R)$. For the summation methods of the class $\mathcal S^\theta$ we find error estimates on $W_s^\mu$. The estimates take into account the behaviour of the Hermite functions both in the oscillatory region and near the turning point: the maxima of functions with different indices are attained at different points, and the supremum of a sum over a block of indices is substantially smaller than the sum of the suprema. Owing to this, in particular, we establish a common lower bound for all finite diagonal methods; for $s\ne4/3$ and $p\ne4/3$ it matches the upper bound and gives the exact order of the error in this class, while at the critical points the gap between the bounds does not exceed $(\log(e/\delta))^{1/4}$ up to a constant. If, in addition, $s>4/3$ or $p<4/3$, then this order is also optimal among all recovery operators that use the whole sequence of data.

Functional electrical stimulation (FES) assists movement, but fixed schedules cannot track how a repeatedly activated muscle changes, and surface electromyography (EMG) near the stimulation electrodes is often obscured by artifacts. Prior work recorded mechanomyography (MMG) passively during closed-loop FES cycling and estimated FES-induced fatigue or predicted fatigue for FES control, without gating stimulation with stimulated-site MMG. In this study, we present VG-FES, a wearable module co-locating microphone-based MMG, programmable charge-balanced biphasic stimulation, and embedded state-gated control at one on-skin thigh site. After each action, we compare rectus femoris (RF) mean power frequency (MPF) with an early-session baseline of the same direction, giving a relative state. A threshold on its window means enables fixed-duration bursts separated by rest periods. During onboard stimulation, MMG remained recordable, whereas near-site EMG was artifact-dominated. Under prolonged external stimulation, cycle-wise MMG tracked an independent local-pressure reference (Spearman rho = 0.968). In leave-one-session-out evaluation on 2284 squat actions (11 sessions, eight participants), MMG-based action segmentation reached an event-detection F1 of 0.923 and direction classification a balanced accuracy of 0.925. Online, over 15.69 min of self-paced squats, one participant received four bursts while MMG acquisition and action monitoring continued. Mechanical feedback from the stimulated site offers wearable FES and assistive robotics a per-action basis for state-gated stimulation.

Traffic sign detection plays a vital role in intelligent transportation systems, supporting the safe operation of advanced driver assistance systems and autonomous vehicles. Despite significant progress in real-time object detection, accurately identifying traffic signs under challenging conditions remains difficult. Variations in lighting, adverse weather, motion blur, and partial occlusion often reduce detection performance, especially for small or visually degraded signs where important spatial information can be lost. To address these challenges, we propose Squeeze-Aware-YOLO, a lightweight traffic sign detection framework that introduces a multi-branch bottleneck module combining convolutional kernels of different sizes with a Squeeze-and-Excitation mechanism to improve feature extraction and adaptive channel weighting. In addition, we integrate the Multi-Branch Auxiliary Feature Pyramid Network and Re-parameterized Heterogeneous Multi-Scale modules to strengthen multi-scale feature representation and feature fusion while preserving real-time inference efficiency for practical deployment. The proposed approach is evaluated on the Bangladeshi Traffic Sign Detection Dataset, the CSUST Chinese Traffic Sign Detection Benchmark 2021, and the German Traffic Sign Detection Benchmark, achieving mAP@0.5 scores of 74.7%, 79.4%, and 84.0%, respectively, while consistently outperforming recent state-of-the-art YOLO-based traffic sign detectors. These findings demonstrate that combining multi-scale receptive fields, channel attention, and enhanced feature fusion enables more accurate and robust traffic sign detection while maintaining the computational efficiency required for real-world deployment.

Materials synthesis planning (MSP) connects AI-driven materials discovery to experimental realization by proposing precursors, synthesis operations, and processing conditions for a target material. Existing approaches predict these components separately or in pairs and treat conditions only as a prediction target. Consequently, MSP remains centered on recipe prediction: it leaves unexplored how conditions can serve as process context for constructing synthesis routes, and it does not address which candidate routes should be executed under limited experimental resources. We propose CondMSP, which extends MSP from recipe prediction to execution-aware planning. CondMSP predicts temperature, time, and atmosphere by combining a joint predictor with variable-specific experts, and then uses these predictions both as process context, grounding the generation and retrieval of operation sequences, and as execution requirements, specifying which thermal stages can share a furnace run and at what cost. This specification is held fixed across all compared methods; within it, CondMSP ranks routes across precursor sets by a common utility and, under furnace budgets, jointly selects routes and allocates their thermal stages by constrained optimization. CondMSP improves condition and operation prediction and, under this shared constraint, raises reference-recipe recovery from 27.9% to 32.9% at three simulated furnace runs per target and from 34.7% to 40.0% at 30 trials, while a new evaluation set of 2,344 targets supports generalization beyond oxide-dominated training chemistry.

Can quantum communication determine the size of an anonymous network when no size bound is known? Classical impossibility proofs rely on the covering lemma: an execution can be reproduced on a larger network with the same local structure. Quantum interference invalidates this argument - a measurement record possible on a network can become impossible on a cover. We show that the impossibility of exact counting nevertheless survives. Our main result is a quantum analogue of the covering lemma that preserves local possibilities rather than complete executions. Within a cyclic covering family, every local measurement outcome possible at a node remains possible at a corresponding node in every sufficiently large network. This holds even after information has travelled around the smaller network, beyond the reach of a simple locality argument. We extend the principle to the universal covering tree and finite covers with sufficiently long cycles. The proof combines tensor-network methods with graph coverings: a common representation across network sizes turns normalization into a constraint on which local outcomes can occur. For uniform synchronous protocols with independently prepared nodes, this support-lifting principle rules out zero-error counting and even the production of a guaranteed finite size bound on anonymous rings and the other cyclic covering families considered here. It also rules out exact leader election with termination in finite time with probability one on these families. Measurements, classical control, and local memory growing with time are allowed; no predetermined running-time bound is assumed. Quantum interference thus defeats execution lifting without enabling infallible certification of network size.

Wearable muscle sensing for rehabilitation, assistive robotics, and electrical-stimulation feedback needs mechanomyography sensors less sensitive to fit changes during wear. In microphone-based mechanomyography, cavity pressure depends on rim contact, a dependence housing design has rarely targeted. This study evaluates a side-vented housing, whose vent opens a pressure-relief path in parallel with the rim to lower contact dependence, and quantifies the acoustic-gain cost. Thermoviscous finite-element simulations compared twenty cavity configurations under prescribed source motion. Paired bench sweeps over 10-200 Hz on both housings compared direct contact with a Gap-ring spacer, a radially grooved spacer ring. At 100 Hz, side venting reduced the simulated acoustic-gain variation across 0.10-0.50 mm rim clearances from 25.8 to 10.0 dB and cost 5.5 dB of gain at 0.20 mm clearance. Velocity fields and branch-flow calculations identified this path. At the higher speaker drive level, the Gap-ring spacer caused median attenuations of 13.6 dB (vented) and 30.1 dB (unvented) relative to direct contact. In this front end, a 35.9-dB signal-gain increase at 100 Hz raised the calculated operational-amplifier voltage-noise gain by 1.6 dB. A single-participant demonstration of paced elbow flexion and extension compared envelope timing and spectra between housings, finding similar envelope timing (correlation 0.935) and broadly similar spectra (correlation 0.956). Vent area trades acoustic gain against contact dependence and can therefore serve as a housing design parameter.

We study constant-round LOCAL approximation for additive local vertex-subset minimization problems. Given a $k$-uniform $\alpha$-approximation on a hereditary source class closed under disjoint union, Bonamy, Das, Gavoille, Picavet, Suomela, and Wesolek obtained ratio $\alpha(d+1)+1$ on locally nice target classes of asymptotic dimension $d$. We prove two transfers with ratios $\alpha+1+\eta$ and $\alpha+\eta$ for every $\eta>0$. These bounds require the stronger local niceness assumption at a larger scale that depends on $\eta$. The first algorithm keeps their repair construction and increases the radius used to detect errors. We average over partitions of the graph to bound the retained cost without the factor $d+1$. These partitions are used only in the analysis. The second algorithm tests several deletion widths around the errors. In each component of the repair zone, it chooses a smallest union of retained choices and an exact repair. Averaging over the widths bounds the cost counted twice and yields ratio $\alpha+\eta$. Both transfers are deterministic and computable, use arbitrary distinct numerical identifiers, and need no knowledge of the graph size. Their outputs also satisfy local uniformity guarantees. For 2-domination on graphs of bounded Euler genus, the same planar source gives ratio $19$ through the earlier transfer and $6+\eta$ through ours. For minimum dominating set, our transfer gives ratio $7+\varepsilon$ on graphs formed by attaching a clique of arbitrary size at one vertex of a planar graph, using the source algorithm of Amiri and Siebertz. This class satisfies our target assumptions and allows nonempty error sets. Its Euler genus is unbounded, so their guarantee for graphs of bounded Euler genus does not cover the whole class.

In real-world indoor fire environments, visual degradation severely impairs instruction grounding, waypoint prediction, and long-horizon exploration in vision-language navigation (VLN). Although image restoration offers a natural means to mitigate such visual degradation, existing methods primarily optimize reconstruction fidelity, which does not necessarily translate into better navigation decisions and may even corrupt task-critical cues when applied indiscriminately. In this paper, we present FireNav, a decision-oriented reconstruction evidence routing framework that aligns visual restoration with navigation utility in VLN. To advance this task, we first contribute FireVLN-Bench, a paired indoor fire navigation benchmark with clean references preserving task semantics, safety supervision, and matched settings for continuous navigation and robot execution with simulated dynamics. Our FireNav retains native visual tokens and treats latent reconstruction features as auxiliary evidence, where a lightweight router leverages visual features, task context, and prior execution states to determine whether such evidence should be admitted and how strongly it should contribute to the visual tokens. To learn when reconstruction is beneficial, we further introduce decision utility calibration, which compares action prediction losses with and without candidate evidence under the same state, thereby supervising evidence selection according to its contribution to waypoint, heading, and arrival prediction. Experiments demonstrate that FireNav consistently outperforms state-of-the-art baselines in both navigation success and path efficiency across evaluation settings.

In practical engineering and scientific computing, a real physical system can be reduced to a boundary value problem, and the adequate boundary information to determine the interior spatial distribution and evolutionary state of the physical field can be solved with linear equations, alternatively, the the real physical problem can be treated with boundary problems with in adequate information problems in field of machine learning. This paper proposes a deep learning framework for the recovery of missing boundary information using an attention-guided graph neural network (GNN) surrogate. The framework extracts features from the physical information on the discretized boundary into a latent space, and the missing boundary quantities are then predicted by the network. Thus, the nonlinear mapping from known boundary quantities to missing boundary quantities is learned using a variational autoencoder (VAE) architecture, while an attention mechanism is incorporated to better model the correlations among different types of boundary information. Furthermore, the interior physical field can be reconstructed using the conventional boundary element method (BEM) integral formulation. The effectiveness of the GNN-BEM framework is evaluated using numerical examples of the two-dimensional Laplace and Poisson equations, which demonstrate good performance in reconstruction accuracy, local detail preservation, and model stability. The results show that GNN-BEM provides a feasible computational framework for boundary information recovery and fast interior physical field reconstruction for partial differential equations under complex boundary conditions.

Many deployed distillation pipelines show the student a truncated view of the teacher: keep the top-k probabilities, renormalize, and minimize a divergence on what remains. Renormalization is assumed to account for the missing mass. We prove that it removes the visibility of tail mass, not the freedom to move it. The renormalized objective depends only on log-odds among the retained tokens, so at each position its gradient along the directions that control tail mass is exactly zero; on a real teacher that component measures $2$-$3\times10^{-15}$ of the gradient. The full-vocabulary gradient equals the truncated gradient plus a residual weighted by the tail masses, and training acts only on the first term. Without the cross-entropy term, a top-k student's tail mass drifts away from the teacher's while a full-vocabulary control moves slightly toward it; with it, the gap shrinks about fourfold but persists. On seven open-weight teachers, the retained set determines the teacher's distribution only from $k=d+1$ ($d$ the hidden dimension): recovery is exact from there on all seven, and commonly cached sizes sit below it. In our off-policy runs the practical effect is real but smaller than the choice of teacher-student pair. Renormalization is a normalization, not a safeguard. The code is available at https://github.com/KurbanIntelligenceLab/top-k-distillation.

World Action Models (WAMs), often built upon pretrained video-generation world models, have emerged as a promising framework for general-purpose and open-ended robot control. However, their real-world deployment often suffers from discontinuous and jerky execution. Existing methods typically address this issue by either improving inference efficiency or enhancing action continuity across prediction chunks, without jointly addressing both challenges. To bridge this gap, we propose StreamingWAM, a streaming-diffusion framework for WAM-based robot control. StreamingWAM combines a noise ladder for chunk-wise action prediction, a prefix rollout module for cross-chunk continuity, and dense video-history memory for temporally consistent conditioning. Experiments on LIBERO, LIBERO-Plus, RoboTwin 2.0, and real-world robotic tasks demonstrate that StreamingWAM improves action continuity across static and dynamic settings, achieving competitive performance while reducing steady-state replanning latency to 72.9 ms in simulation and 57.3 ms on the real robot. The method also produces smoother action transitions, while the streaming staircase reduces sensitivity to the execution horizon K.

We study the problem of inverse garment design, reconstructing simulation-ready 3D outfits with arbitrary layers from a single image, with visual agents. Unlike differentiable optimization that is time-consuming, or feed-forward methods limited by training data, we introduce DressAgent, a hierarchical agentic system that recovers garment sewing patterns through planning and closed-loop refinement grounded in physical simulation. To this end, we first present an enhanced representation based on GarmentCode for diverse styles and arbitrary dressing layers, and build the first multi-layer sewing pattern dataset with over 15K outfits for supervision and benchmarking. Inverse design then proceeds coarse to fine: In layer planning and coarse estimation, agents infer layer order and relations, peel each layer into its own view, and estimate coarse patterns. In fine-grained refinement, an agentic trio of Tailor, Critic, and Orchestrator refines patterns in a closed loop of editing, simulating, and judging against the reference with visual and measured evidence. Carefully designed tools and an evolving memory make this refinement converge on the target geometry. We also design an agentic generation method for high-fidelity textures. DressAgent outperforms feed-forward methods by a large margin, surpasses or matches differentiable optimization in a fraction of the time, and generalizes to complex layering, diverse poses, and body shapes.

Immersive extended reality (XR) applications are key enablers of human-centric connectivity in next-generation networks. However, meeting their high-throughput, ultra-low-latency, and high-reliability demands while preventing cyber-sickness remains a critical challenge, especially in environments with non-deterministic wireless connectivity. Fiber-To-The-Room (FTTR), extending widely deployed Fiber-To-The-Home (FTTH) infrastructure, offers a promising solution. Yet, satisfying the stringent Quality-of-Experience (QoE) requirements of XR collaborators demands robust resource allocation across the multi-tier cascade of FTTH, FTTR, and WiFi segments. In this paper, we propose a coarse-grained traffic forecasting-empowered proactive resource allocation and intelligent in-premise roaming framework. Here, artificial intelligence (AI) model complexities are scaled to the computational capabilities of each network segment. At the FTTR segment, we employ a hybrid temporal convolutional and bi-directional long short-term memory (TCN-BiLSTM) network for multi-modal XR traffic forecasting, integrating passive optical network (PON) and WiFi access points with multi-link operation (MLO) and orthogonal frequency-division multiple access (OFDMA). The resulting long-horizon forecasts reduce inference frequency, enabling coordinated FTTR-WiFi resource allocation and seamless network-orchestrated mobility across rooms. At the FTTH segment, we deploy Transformer-based time-series models (e.g., Informer, Autoformer, PatchTST, and TFT) to predict the complex XR traffic aggregated in FTTR segments for transmission over the FTTH backhaul. Evaluations show Transformer-based models are up to $\sim$75% more accurate in long-horizon forecasting than the hybrid model. Extensive simulations confirm the proposed framework significantly outperforms traditional methods in satisfying XR QoE requirements.

This paper presents an algorithm for detecting and fitting multiple ellipsoids in 2D, 3D and higher-dimensional data. The representation is based on tangents to the ellipsoid together with their contact points, rather than on points alone. A tangent contributes d independent equations where a point contributes one, so each hypothesis needs fewer samples, which in turn reduces the number of iterations required to recover every ellipsoid. The problem is posed as sets of linear equations whose solutions – the ellipsoid parameters – are recovered robustly by clustering in parameter space. The algorithm works in all dimensions and outperforms the state of the art in 2D. The approach applies in general to the robust solution of linear equations, and so has utility beyond ellipsoid fitting.

Disaggregated LLM inference executes the prefill and decode phases on separate resource pools, reducing interference between the two phases. However, this architecture introduces a coupled scheduling problem because a request that completes prefill retains its prompt KV cache in prefill memory until it is admitted to the decode pool. Prefill and decode admission decisions therefore jointly determine memory occupancy and request completion times. To the best of our knowledge, this problem lacks a formal formulation, and no approximation guarantees are known for request scheduling in disaggregated LLM inference. We formalize this setting as the dual-memory request scheduling (DMRS) problem, whose objective is to minimize the sum of request completion times, and prove that it is strongly NP-hard. We develop two deterministic polynomial-time approximation algorithms. The first applies to instances satisfying $\alpha<1$ and $\beta<1$ and achieves an approximation ratio of $8+\frac{4}{1-\alpha}+\frac{8}{1-\beta}$, where $\alpha$ and $\beta$ are the maximum normalized prefill and peak decode memory demands among all requests, respectively. In particular, it yields a $32$-approximation when $\alpha,\beta\leq 1/2$. The second algorithm applies to all feasible instances and achieves a uniform $36$-approximation. Finally, evaluations on two H100 GPUs using trace-derived offline workloads show that the proposed algorithms reduce the sum of request completion times relative to multiple vLLM deployment baselines.

Post-training expert pruning is a promising deployment technique for Mixture-of-Experts (MoE) models. The calibration set is an integral part of the pruning process, however, its correlation with evaluation performance is largely misunderstood and unintuitive. In this work, we present an empirical study of the robustness of calibration perturbations, specifically in MoE models. Showing how calibration size, language, and data distribution affect layer-wise expert pruning for gpt-oss-20b. Our findings reveal that (1) domain-general pruning quality receives minimal benefit from scaling calibration data beyond small sample sizes, (2) cross-lingual calibration transfers effectively, with English-only calibration maintaining strong performance across seven languages of varying typologies and scripts, and (3) a diverse calibration data resembling the pretraining distribution can yield comparable or superior performance to test-distribution data. Ultimately, we provide practical insights as a recipe for better understanding calibration for more effective domain-general expert pruning.