Science & Mathematics

Structure-from-motion (SfM) from the dual-fisheye video of a consumer 360-degree camera recovers geometry only up to a similarity transform; in the pipeline considered here, metric scale is obtained from smartphone ARKit or ARCore poses captured alongside the camera. This paper evaluates a self-contained alternative: the IMU embedded in the Insta360 X5 and X6, recorded at 1 kHz. The sensor is characterised (noise density, camera-IMU synchronisation, accelerometer scale factor, bias stability), the camera-IMU rotation and frame timing are estimated from the data, and the scale is recovered by a linear visual-inertial alignment over the SfM poses of a whole survey. The standard formulation, with gravity and accelerometer bias shared across the survey, overestimates the scale by 3.4% on average; estimating each contiguous segment independently and averaging removes the bias. On nine surveys with a co-mounted LiDAR scanner the inertial scale deviates from the LiDAR by 0.74% in mean absolute error, against 1.94% for the smartphone scale; on printed fiducial markers both agree within the 1% uncertainty of the target. On the reference trajectories of thirty sequences of a public benchmark, brought to scales of 1/40, 1 and 40, the estimator recovers the metric scale with a mean error of -0.07% and a standard deviation of 1.28%. On fifty production surveys where the smartphone is the only reference, the two scales agree to a standard deviation of 2.6%, the inertial scale being 0.9% lower on average; the difference does not depend on window construction, grows with the extent of the smartphone anchor, and cannot be attributed to either reference. An ablation over window construction and a comparison with monocular metric-depth networks complete the study.

Looped Transformers increase computation per token by repeatedly applying a shared backbone, offering a way to improve language modeling and downstream task performance at a fixed backbone parameter count. However, each loop depends on the preceding loop, so increasing recurrent depth also increases decoding latency. Existing approaches include Parallel Loop Transformers, which reduce latency by overlapping loop computation across token positions, but yield rapidly diminishing marginal accuracy gains at larger depths. To address this challenge, we introduce LoopFold, which folds computation at multiple recurrent depths into parallel batches for the same token prefix. A lightweight predictor estimates inputs to later loops so they can start before earlier loops finish. A corrector then refines the outputs of these later loops using newly available results from earlier loops. All components are trained jointly using only the final next-token prediction loss. In experiments with 1.4B models, deeper LoopFold configurations achieve progressively higher average zero-shot accuracy across eight tasks, rising from 53.41% to 56.38%. Vanilla looped models and PLT show limited gains over the same depth range. Compared with a nine-loop vanilla model, LoopFold uses five sequential stages and additional parallel computation to reduce decoding latency by 46%.

Language models are increasingly considered for clinical decision support, where their answers should be grounded in clinical evidence and remain robust to irrelevant information. Prior work shows that a single irrelevant sentence can lower the accuracy of medical models, but not where inside the model its effect can be removed, or how to do so without knowing which sentence it is. We find that blocking attention to a known distractor restores the answer, whereas head masking can suppress correct answers. We also show that the model's hidden states mark the distractor as the sentence that least fits the case even when the model follows it. To remove a distractor's effect without knowing which sentence it is, we propose EXCISE, an inference-time method that blocks the sentence most influencing the answer, but only if the model's hidden states mark it as inconsistent with the case. On MedQA-TD (MedQA with Targeted Distractors), a benchmark we construct in which each distractor targets a specific wrong option, EXCISE gains 16.7-22.8 accuracy points at a 1.5-2.1-point cost on the same questions without distractors, outperforming prompting, rewriting, and steering baselines for all three models. Its gains extend to MedDistractQA and BiasMedQA, as well as chain-of-thought generation.

Multi-fidelity training, which combines abundant low-fidelity data with scarce high-fidelity labels, is increasingly common in atomistic modeling, but its effect on predictive uncertainty remains largely unexplored. We investigate this question using deep ensembles of mean-variance MACE models trained on ANI-1x, contrasting high-fidelity training from scratch with low-fidelity pretraining followed by fine-tuning. We characterize aleatoric, epistemic, and total uncertainty throughout training and under two distribution shifts: unseen molecular systems and higher-energy configurations. We find that pretraining reduces prediction error and improves uncertainty-based ranking across all test sets, but does not yield well-calibrated predictive variances; calibration deteriorates more severely under energy extrapolation. Training-dynamics analysis localizes most of this calibration degradation to high-fidelity optimization, during which the learned aleatoric variance becomes overconfident on held-out data. Moreover, multi-fidelity fine-tuning produces a more effective uncertainty-based acquisition function than high-fidelity training from scratch. These findings show that uncertainty ranking and calibration capture distinct properties and must be assessed separately in multi-fidelity atomistic models, especially under extrapolation.

People who are blind or have low vision (BLV) are largely excluded from 3D modeling tools that rely on visual representations of geometry. Natural language descriptions offer a potential path to accessibility, but it is unclear whether language alone can support the deep spatial understanding that editing, not just browsing, requires. Advances in generative AI offer the promise of automatic model description, yet descriptions suited to BLV modelers are structurally distinct from alt-text and from existing 3D model captioning datasets, and the domain-specific requirements for generating them are not well understood. We address this through two coordinated studies: an autoethnographic pilot in which the blind first author worked with 12 sighted participants to develop a taxonomy of descriptive strategies, and a Wizard-of-Oz main study that applied those strategies as idealized scripts with 10 BLV participants. We identify a systematic use-to-editing gap: descriptions that support accurate but incomplete mental models are insufficient for specifying modifications, because editing requires knowing the state of features the user did not know to ask about. We show that this gap produces four failure types (accuracy, completeness, expression, and navigation) that accuracy-focused evaluation frameworks cannot detect. We propose a two-phase evaluation framework that diagnoses the failure type a description system exhibits and guides developers toward targeted design interventions, thereby providing a methodological standard for assessing whether description-based systems genuinely support BLV modelers as creators.

Experts can master fine hand-object manipulation skills, yet limited conscious perception of rapid tactile interactions hinders their ability to articulate the underlying control mechanisms. Badminton racket control requires rapid, fine hand-force coordination, providing a demanding testbed for uncovering these mechanisms. An 864-node tactile-inertial glove with synchronized 100-Hz acquisition recorded over 14,000 strokes from more than 70 participants across five skills. Inspired by sensorimotor control, our mechanics-grounded two-layer framework inferred object-driving and grip-stabilizing control from tactile history, while resolving their spatiotemporal force realization. We found that higher proficiency was associated with selective rather than extensive control: greater skill specificity, anticipatory regulation, focal distal modulation and lower geometric torque cancellation, while spatiotemporal coordination showed a non-monotonic acquire-refine pattern. We translated these mechanisms into personalized guidance targeting force control and force-motion coordination. In a randomized cohort, mechanism-guided learners gained 0.72 expert-rated skill levels versus 0.05 with level-only feedback, making control mechanisms actionable.

Serverless computing platforms must preserve quality of service (QoS) while adapting to highly bursty and unpredictable workloads. To cope with workload fluctuations while satisfying QoS requirements and service-level agreements (SLAs), these platforms rely on automatic resource scaling. Commercial function-as-a-service (FaaS) platforms employ different autoscaling strategies to dynamically adjust the number of active function instances and the available processing capacity. In this work, three representative autoscaling strategies are modeled and compared: scale-per-request (SPR), concurrencylevel scaling (CLS), and metric-based scaling (MBS). To enable systematic evaluation independent of proprietary platform-specific implementations, we develop a unified modeling framework that combines analytical queueing network (QN) modeling with queueing Petri nets (QPN) implemented in Java Modeling Tools (JMT) to evaluate autoscaling behavior under two-phase hyper-exponential (H2) arrivals. Evaluation is conducted using four system-level metrics: throughput, mean response time, system population, and system power. The framework captures dynamic function instance (FI) activation, concurrency constraints, and queue buildup under saturation. Comparative analysis reveals a clear trade-off between peak performance and stability: SPR provides the lowest delay under moderate load but becomes fragile near capacity limits; CLS is governed by concurrency tokens and is highly sensitive to configuration under bursty conditions; whereas MBS provides the most stable QoS across the workload range by keeping utilization below a critical region and preventing explosive queue growth under critical load.

Image editors built on diffusion transformers (DiTs), such as Flux Kontext, make it possible to turn a single photograph into a convincing manipulated image, including harmful and non-consensual edits. Proactive defenses aim to prevent such manipulations by adding an imperceptible perturbation before an image is shared. Existing defenses for DiT-based in-context editors make the output attend less to the protected reference image. But this leaves the model free to follow the edit instruction using the text and its own generated tokens, often producing a new, weakly grounded image rather than preventing the edit. We find that the opposite strategy is more effective: instead of leading the model to ignore the reference, we encourage it to copy it. Through controlled attention interventions, we show that concentrating target-token attention on the reference image in only a few early blocks is sufficient to collapse the editing behavior, causing the model to reproduce its input, often as a lossy copy, rather than execute the requested transformation. Based on this insight, we introduce Copying-as-Defense (CaD), which maximizes per-token target-to-reference attention mass in early transformer blocks. Since attention rows are normalized, increasing reference attention suppresses the attention allocated to the text and generated canvas. Restricting the objective to early blocks also lets us truncate the backpropagation path, keeping optimization cheap. Across editors and benchmarks, CaD reduces average edit success from 81.4% on unprotected images to 24.0%, a 70.6% relative reduction, and by a further 44.3% relative to DeContext, the strongest baseline (43.0%). CaD is also roughly 1.8-2.0x faster per optimization step than comparable methods.

We introduce ExCal, a normal-only multivariate time series anomaly detector that combines raw state retrieval, a signal/derivative subspace, and robust statistics across temporal scales. Each expert map receives a high normality floor from the training prefix. During scoring, positive excess is retained, attenuated for broadly firing maps, and combined with a small rank term. The method fits reference memories and projections without gradient optimization or anomaly labels. Prevalence and ranks depend on the complete scored sequence, making inference batch and transductive rather than online. On 180 series from 17 TSB-AD-M families, ExCal obtains AUC-PR $0.421\pm0.37$ and VUS-PR $0.452\pm0.37$ over three seeds, ranking second among 24 methods. Holm-adjusted paired tests detect no significant difference from AxonAD or StreamVAE, which are the two leading non-pretrained methods on TSB-AD-M, on either primary metric, while finding significant positive differences against the other 21 baselines. A separate single-run probe over three series measures median model-wrapper time of $0.0662$ s. Ablations associate the largest primary-metric changes with raw retrieval and excess-magnitude preservation. The evaluated score is decomposable by expert, but its diagnostic utility and online suitability remain untested.

By virtue of the sheer compute demands of large-scale artificial intelligence (AI) and high-performance computing (HPC), modern multi-core microprocessors suffer from severe power density limits and thermal throttling, entailing an urgent need for principled methods for predictive, physics-informed dynamical thermal management (DTM). Here, we introduce a hardware/software co-design physics-informed PDE-constrained DTM throttling optimal control algorithm, ADACT (Adjoint-Driven Adaptive Co-Design Thermal Control), that minimizes thermal risk and cooling energy overhead while maximizing core execution throughput. The underlying physical temperature field is governed by a two-dimensional nonhomogeneous parabolic reaction-diffusion equation with spatially localized, time-varying power source terms and variable convective cooling subject to nonhomogeneous Dirichlet boundary conditions. We employ a discretization of the continuous state PDE via a Crank-Nicolson finite difference scheme with a 5-point discrete Laplacian operator on a uniform spatial mesh, which preserves unconditional numerical stability and $O(Δx^2 + Δy^2 + Δt^2)$ second-order temporal and spatial accuracy. To efficiently compute exact gradients, we construct and implement a continuous backward-in-time adjoint state method with L-BFGS as our optimization algorithm of choice. The accuracy and second-order convergence rates of both forward state and backward adjoint solvers are quantitatively verified via the Method of Manufactured Solutions (MMS), and gradient accuracy is validated using directional derivative Taylor tests. Numerical experiments demonstrate that ADACT dynamically throttles software power workloads and modulates hardware cooling efforts to strictly uphold physical thermal thresholds while maintaining operational throughput.

We present the Star-formation from Photometry through the Addition of Medium-bands (SPAM) survey, a 62.9 hr JWST/NIRCam imaging program to add ten new filters, nine medium-band filters, and one broad-band filter to the CEERS legacy dataset. We describe the observation strategy, present the data, and detail the data-reduction process. We include a complete reduction of imaging in the legacy CEERS footprint, inclusive of all publicly available photometric parallel and GO observations within the legacy CEERS footprint. We present a multi-band photometric and photometric redshift catalog which includes all medium- and broad-band filters available on NIRCam. We find that adding medium bands from the SPAM and MINERVA programs improves photometric redshifts. We discuss how these data advance the SPAM survey's key science goals: better-constrained photometric redshifts, identification of high-redshift overdensity members, improved recovery of emission-line strengths, and more accurate star formation histories.

Long-horizon embodied navigation may be interrupted by a higher-priority instruction before the current task sequence is complete. The agent must finish its current atomic task, execute the inserted task, and resume the remaining sequence from the updated environment state, where its viewpoint and available navigation candidates may have changed. To evaluate this capability, we introduce PIFR benchmark, a novel task for runtime task preemption and resumption in embodied navigation. PIFR includes single, repeated, and nested interruption settings,with stage-specific metrics for inserted-task completion, resumption of the original task, and completion of the full sequence. Evaluations show that recovery performance varies across interruption conditions, while aggregate success can obscure differences in interruption handling and resumption. We further propose a lightweight candidate-ranking baseline to improve target re-grounding after interruption. Together, PIFR and the baseline provide a systematic framework for studying interruption handling and task recovery in long-horizon embodied navigation.

We present the Infinity Machine, a self-improving AI agent that we use for research on problems whose answers aprogram or data can check. Each candidate solution is produced by a full tool-using agent that works for up to an hour in a fresh sandbox; proposers see only a public score, and a private split decides what counts. The harness itself(prompts, tools, skills, playbook, meta agent) is an editable genome, changed only through a kernel-enforced gate with held-out and cost checks. An autonomous mode goes from a plain-text goal to a verified report with no human step: an agent picks an open problem and writes its verifier, the kernel calibrates it, an adversarial agent audits it, and an agent that never sees the verifier re-checks the result. Across fourteen problems in six days, for about $180 of model usage, the system produced four results beyond the published state of the art on catalogued open problems in quantum information: a certified depolarizing-channel threshold point p = 0.064966, a qutrit positivity frontier of 0.2910672, a 44-gate depth-6 Golay state preparation, and a stabilizer-rank bound chi(T^7)

Linear hyperspectral unmixing aims at retrieving pure material spectra present in an observed scene and their associated abundance fractions in each pixel. In most practical situations, hyperspectral images are affected by spectral variability (SV) also known as intra-class variability, which is caused by different physical phenomena including atmospheric effects, intrinsic variability of the pure materials and illumination or topographic changes. Nevertheless, the standard hyperspectral unmixing methods that rely on the Linear Mixing Model (LMM) do not consider these phenomena. Thus, neglecting them will spread errors throughout the linear hyperspectral unmixing process. To avoid this problem, this paper proposes two hyperspectral unmixing algorithms dedicated to a parametric model called Structured Additive Variability LMM (SAV-LMM). This mixing model handles SV in an additive manner by incorporating two terms into the standard LMM. Moreover, the designed algorithms are based on Nonnegative Matrix Factorization (NMF) and optimize a cost function which integrates additional penalization accounting for SV. Furthermore, a novel improved initialization scheme, which takes SV into account, is proposed. Experiments conducted on both synthetic and real datasets show the effectiveness of the proposed algorithms for linearly unmixing hyperspectral images with SV compared to the tested state-of-the-art methods.

A full-duplex spoken dialogue model must yield when the user cuts in and hold when the user merely backchannels. Sequence-level RL with a timing reward (GRPO) is the usual recipe; we test its whole-rollout form on four open models and find two failures. A binary window reward gives zero advantage on exactly the windows that most need learning (a zero-gradient trap), and the one score per rollout is spread over listening-state decisions the reward is not written in terms of (credit leakage), so the model talks into silence or stops answering. Both come from sampling a reward that is a known function of a few binary control decisions. We propose \emph{window likelihood} (\WL): sample a continuation per window, but integrate in closed form over the stop decisions it visits and maximise the log-probability of acting correctly---a noisy-OR for yielding, a survival product for holding. On HumDial-FDBench, \WL\ gives the best yield/hold F1 on all six structurally different open models we post-train, exceeds Kyutai's released GRPO checkpoints (trained on different data), and adds no re-opening of the floor on any model.