arXiv:2603.28896v2 Announce Type: replace Abstract: Feed-forward foundation models for multi-view 3-dimensional (3D) reconstruction have been trained on large-scale datasets of perspective images; when tested on wide field-of-view images, e.g., from a fisheye camera, their performance degrades. This degradation arises from changes in spatial arrangements of pixels induced by the non-linear projection model that maps 3D points onto the 2D image plane. While one may surmise that training on fisheye images would resolve this problem, there are far fewer fisheye images with ground truth than perspective images, which limits generalization. To enable inference on imagery exhibiting high radial distortion, we propose \textit{Fisheye3R}, a novel adaptation framework that extends these multi-view 3D reconstruction foundation models to natively accommodate fisheye inputs without performance regression on perspective images. To address the scarcity of fisheye images and ground truth, we introduce flexible learning schemes that support self-supervised adaptation using only unlabeled perspective images and supervised adaptation without any fisheye training data. Extensive experiments across three foundation models, including VGGT, $\pi^3$, and MapAnything, demonstrate that our approach consistently improves camera pose, depth, point map, and field-of-view estimation on fisheye images. Code is available at https://github.com/android-xr/fisheye3r.
Science Journals
arXiv:2607.02110v1 Announce Type: new Abstract: We investigate the epistemic opacity of computer simulations and machine learning methods in the context of black hole imaging. We argue that there are forms of opacity-including opacity resulting from the use of machine learning-which do not need to affect the reliability of an inference when it is seen as a part of a broader inferential framework. We propose conditions under which that can plausibly be the case, and discuss how opaque methods can be useful in the context of the (next generation) Event Horizon Telescope. However, we also argue that at least one problematic form of opacity is currently present in black hole imaging: GRMHD models of Sagittarius A* are opaque. This form of opacity signals the limitations of current understanding of the models of this source, and constrains the potential uses of ML models in future observations.
arXiv:2607.02347v1 Announce Type: cross Abstract: Variations in solar radius (hereafter R_Sun) is a key reference for solar magnetic activity in time. The sunlight amount may have varied with R_Sun and had an effect on the Earth's climate in the past. Eclipse observations offer a unique opportunity to measure the absolute R_Sun value before modern direct observations. The scientific community has discussed a possible long-term R_Sun variability from 1715 onward. Prior to their coverage, Clavius' eclipse reports had been subjected to qualitative debates regarding the local eclipse visibility and a possible secular R_Sun trend. This study leverages the recent dramatic developments of lunar topography data and ephemeris data to provide an effective resolution of this debate. Clavius' eclipse reports described an explicit totality in 1560 at Coimbra and a "slender circle" around the eclipsing Moon in 1567 at Rome. Our study revised the {\Delta}T constraints of -492 s =< {\Delta}T =< 200 s in 1560 and 140 s =< {\Delta}T =< 151 s in 1567 to satisfy Clavius' descriptions, considering the lunar limb profile and assuming Auwers' canonical R_Sun. This study constrains the R_Sun margin of 1567, utilising three scenarios to interpret Clavius' account. The local totality requires an upper R_Sun limit of 1567 as R_Sun =< 696200 km in absolute size (959.92" in angular size), indicating no linear secular R_Sun shrinkage but possible R_Sun oscillations on a centennial timescale. Conversely, the annularity scenario is considered unlikely because it requires an R_Sun decrease of 7.5" within 3 centuries, even beyond the capacity of extreme shrinking-Sun hypotheses.
arXiv:2607.01919v1 Announce Type: new Abstract: Agentic systems enhance their capabilities by invoking external tools and maintaining persistent memory. However, these external dependencies introduce novel attack surfaces. Recent tool and memory poisoning attacks show that maliciously crafted tool descriptors and poisoned memory can covertly bias agent behavior. These threats reflect a deeper issue: the lack of verifiable continuity in the agent's contextual state for planning and execution. We present ElephantAgent, a protocol that enforces Contextual State Continuity to defend against contextual state poisoning. Inspired by prior state-continuity mechanisms (e.g., Nimble), ElephantAgent extends this protection to the evolving contextual state of agentic systems. We define the contextual state as the bounded, security-critical subset of the agent's entire context (e.g., tool state and memory). Before processing each query, ElephantAgent recomputes the digest of the local contextual state and verifies it against the latest authorized digest. Using replicated trusted hardware, ElephantAgent maintains a linearizable ledger of authorized contextual state transitions and detects out-of-band state tampering. To handle in-band semantic abuse, ElephantAgent additionally provides Historical Traceability, enabling conditional post-hoc audit and recovery to a known-good prior state.
arXiv:2607.02163v1 Announce Type: new Abstract: Exploring similar nodes in attributed networks represents a key challenge in data mining. While recent representation learning methods embed networks into low-dimensional vectors, they often implicitly assume a uniform and continuous feature space. This paper proposes a visual analytics approach using dimensionality reduction to help clarify the true topological structure of high-dimensional feature spaces formed by nodes' neighborhood attribute profiles. Analyzing inter-firm transaction networks indicates that structural roles can form complex, non-linear manifolds with density biases. Comparing this feature space with industry classifications suggested: (1) supply chain hierarchies transition continuously; (2) categories treated identically under general semantics can be clearly separated by actual transaction networks; and (3) a single industry label may fragment into multiple regions. These findings suggest potential limitations in assuming identical semantics imply similar structural roles and highlight the possible need for new similarity metrics aligned with manifold topology.
arXiv:2607.02187v1 Announce Type: new Abstract: Distributed machine learning enables collaborative model training without centralizing data, but it also exposes learning processes to privacy leakage and malicious manipulation. Existing defenses typically address these threats in isolation and are often tailored to specific learning paradigms or model architectures, limiting their applicability in realistic deployments. In particular, federated learning and decentralized learning exhibit distinct adversarial surfaces that are rarely addressed within a unified framework. In this paper, we present a model-agnostic framework for adversary-resistant distributed learning that jointly addresses privacy preservation and malicious behavior across both federated and decentralized settings. Our approach combines paradigm-specific defense mechanisms with GPBACC, a privacy-enhancing coded computing technique applicable to arbitrary machine learning models. For federated learning, we integrate robust aggregation strategies to mitigate the impact of malicious participants, while for decentralized learning we employ approximate decode-and-compare and group testing techniques to enable lightweight verification and adversary isolation without relying on a trusted aggregator. Crucially, we evaluate the proposed framework through an explicit, attack-driven analysis. We implement representative privacy attacks and malicious behaviors, and empirically demonstrate that the combination of GPBACC with robust aggregation and verification mechanisms significantly reduces privacy leakage and improves resilience against active adversaries. These results suggest that privacy-enhancing coded computing, when combined with appropriate adversary-resistance strategies, provides a practical and deployable foundation for secure distributed machine learning.
arXiv:2603.29466v2 Announce Type: replace Abstract: Existing methods for quantifying predictive uncertainty in neural networks are either computationally intractable for large language models or require access to training data that is typically unavailable. We derive a lightweight alternative through two approximations: a first-order Taylor expansion that expresses uncertainty in terms of the gradient of the prediction and the parameter covariance, and an isotropy assumption on the parameter covariance. Together, these yield epistemic uncertainty as the squared gradient norm and aleatoric uncertainty as the Bernoulli variance of the point prediction, from a single forward-backward pass through an unmodified pretrained model. We justify the isotropy assumption by showing that covariance estimates built from non-training data introduce structured distortions that isotropic covariance avoids, and that theoretical results on the spectral properties of large networks support the approximation at scale. Validation against reference Markov Chain Monte Carlo estimates on synthetic problems shows strong correspondence that improves with model size. We then use the estimates to investigate when each uncertainty type carries useful signal for predicting answer correctness in question answering with large language models, revealing a benchmark-dependent divergence: the combined estimate achieves the highest mean AUROC on TruthfulQA, where questions involve genuine conflict between plausible answers, but falls to near chance on TriviaQA's factual recall, suggesting that parameter-level uncertainty captures a fundamentally different signal than self-assessment methods.
arXiv:2607.02198v1 Announce Type: new Abstract: Human-AI teaming has received increasing attention in the literature. However, the range of studies conducted in multiple domains make it difficult to understand what types of teams are being studied, and in what ways are they similar/different from one another. In this study, we analyse 53 papers on human-AI teams and categorise them into five main clusters based on psychological taxonomies of teaming; AI Assistant, Ad-hoc Dependency, Ad-hoc Forced Dependency, Paired Equanimity, and Group Equanimity. Each cluster represents a unique combination of holistic team-level characteristics, indicating there are multiple disparate team types studied under the same definition. In turn, this raises the question of whether insights are truly transferable between papers. We conclude with guidance on how to identify the types of human-AI teams studied, a checklist for reporting a human-AI team in research work, and ways in which the field can be further synthesised.
arXiv:2607.02213v1 Announce Type: new Abstract: Traditionally, midlatitude storm tracks are viewed as being driven by meridional temperature gradients maintained by differential solar heating. Yet in the Southern Hemisphere, storm activity remains strong even when the summertime insolation gradient nearly vanishes. Here, we show that storm-track cloud radiative effects play a major role in maintaining the Southern Hemisphere storm activity. Satellite observations reveal that sunlight reflected by midlatitude clouds in early summer creates a substantial meridional gradient in surface heating, despite the nearly uniform summer insolation. Idealized aquaplanet simulations then show that shortwave cloud radiative effects reinforce meridional sea-surface temperature gradients, thereby strengthening storm activity primarily during late summer and autumn, while longwave cloud effects partly offset this response. To interpret these results, we develop a simple theoretical model linking storms, clouds, and sea-surface temperature gradients. The model reproduces the simulated seasonal response and identifies two emergent cloud properties that control the feedback strength: the maximum attainable cloud albedo and the sensitivity of cloud cover to storm activity. Together, these findings indicate that cloud radiative feedbacks are key to maintaining the thermal gradients that sustain storm activity. More broadly, they reveal a strong coupling among storms, clouds, and the ocean spanning distinct spatial and temporal scales.
arXiv:2603.29616v2 Announce Type: replace Abstract: The inherent complexity of video understanding makes it difficult to determine whether Video-LLM benchmark performance stems from visual perception, linguistic reasoning, or knowledge priors. While many benchmarks have emerged to assess high-level reasoning, shared criteria for evaluating video understanding remain largely overlooked. Instead of introducing yet another benchmark, we take a step back to re-examine the criteria for evaluating video understanding. In this work, we introduce Video-Oasis, a sustainable diagnostic suite for systematically auditing existing video understanding benchmarks. This audit reveals that 55\% of existing benchmark samples are solvable without visual input or temporal context. After filtering these shortcuts, the remaining video-native challenges expose a substantial capability gap: state-of-the-art models perform only marginally above random guessing. Building on these findings, we use the distilled challenges as a testbed to investigate which algorithmic design choices contribute to robust video understanding. We hope our work provides a practical foundation for constructing rigorous video benchmarks and evaluating future Video-LLMs. Code is available at https://github.com/sejong-rcv/Video-Oasis.
arXiv:2512.11891v2 Announce Type: replace Abstract: Vision-Language-Action (VLA) models have demonstrated remarkable capabilities in generalizing across diverse robotic manipulation tasks. However, deploying these models in unstructured environments remains challenging due to the critical need for simultaneous task compliance and safety assurance, particularly in preventing potential collisions during physical interactions. In this work, we introduce a Vision-Language-Safe Action (VLSA) architecture, named AEGIS, which contains a plug-and-play safety constraint (SC) layer formulated via control barrier functions. AEGIS integrates directly with existing VLA models to improve safety with theoretical guarantees, while maintaining their original instruction-following performance. To evaluate the efficacy of our architecture, we construct a comprehensive safety-critical benchmark SafeLIBERO, spanning distinct manipulation scenarios characterized by varying degrees of spatial complexity and obstacle intervention. Extensive experiments demonstrate the superiority of our method over state-of-the-art baselines. Notably, AEGIS achieves over 50% improvement in obstacle avoidance rate while substantially increasing the task success rate by nearly 10%. All benchmark datasets, code, and supplementary materials are publicly available at https://vlsa-aegis.github.io/.
arXiv:2607.02283v1 Announce Type: new Abstract: In-context learning (ICL) operates via implicit gradient descent embedded in the forward pass of modern AI architectures -- Transformers, Mamba, state-space models, and MLPs. Capturing this capability in biologically plausible Spiking Neural Networks (SNNs) has remained an open challenge: existing SNNs fail the Garg-2022 benchmark at non-trivial task dimensions. We trace this failure to a structural assumption: prior SNN designs route adaptation through inference-time synaptic plasticity, viewing the dendritic compartment as a passive conduit for error or teacher signals. We challenge this assumption. The subthreshold dynamics of a single dendritic compartment already implement a complete online learning algorithm. By treating the compartment as the computational substrate rather than a passive conduit, we propose DendriCL -- a single-layer compartmental spiking architecture whose apical recurrence is structurally identical to leaky online Widrow-Hoff LMS. This dynamics-only update collapses the architectural depth required for general-purpose ICL to a single layer. DendriCL is uniquely seed-stable at super-dimensional Garg-2022 ICL -- where dense Transformers exhibit grokking-style instability and fail past moderate task dimension -- and a linear probe recovers the reference online-LMS trajectory directly from the apical membrane at R^2 = 0.93, showing the algorithm is structurally embedded in the dynamics rather than implicitly discovered during training. Taken together, ICL requires neither attention, depth, nor inference-time plasticity: a single compartment with online-LMS dynamics is sufficient.
arXiv:2607.02336v1 Announce Type: new Abstract: The Shifted Boundary Method (SBM) sidesteps body-fitted meshing by shifting boundary conditions onto a surrogate boundary and correcting for the displacement through Taylor expansions. Despite its broad analysis and application, scalable iterative solvers for the incompressible Stokes equations remain underdeveloped. We present a block preconditioner for SBM--Stokes discretisations that uses the velocity block together with a pressure mass matrix as a Schur complement approximation. Because the SBM system is non-symmetric, classical operator preconditioning does not apply directly; a field-of-values analysis instead shows that the non-symmetric SBM contributions act as asymptotically small perturbations of a standard saddle-point operator, yielding mesh-independent GMRES convergence on sufficiently fine meshes. Numerical experiments demonstrate iteration counts under refinement across geometries of increasing complexity. We expose a coarse-mesh regime in which an under-resolved grid produces elevated iteration counts, an artefact of insufficient resolution that vanishes once the mesh captures the geometry.
arXiv:2607.02404v1 Announce Type: new Abstract: Image encoders trained with LeJEPA can deliver strong features for downstream tasks, but, like other image-level self-supervised methods, typically require large training datasets. Aligning representations at the level of objects rather than whole scenes promises greater data efficiency, but doing this in a completely self-supervised way, effectively jointly partitioning a scene and representing its objects, is unstable: the two are locked in a cyclic dependency, partitioning requires meaningful representations, while meaningful representations require consistent partitioning. We sidestep this instability by taking object masks as given during training, using cheap, off-the-shelf SAM proposals. We extend LeJEPA - whose distributional anti-collapse objective ports naturally from whole images to variable-sized sets of objects - to align object-centric representations rather than whole images. An additional instance-separating loss, which treats other objects in the same scene as negatives, further boosts downstream performance. Across two model scales and 10-100% of COCO, object-level LeJEPA outperforms image-level LeJEPA on tracking (DAVIS), classification (ImageNet-1k), segmentation (ADE20k), and re-identification (NAVI).
arXiv:2604.01761v2 Announce Type: replace Abstract: Video diffusion models have recently been applied with success to problems in content generation, novel view synthesis, and, more broadly, world simulation. Many applications in generation and transfer rely on conditioning these models, typically through perceptual, geometric, or simple semantic signals, fundamentally using them as generative renderers. At the same time, high-dimensional features obtained from large-scale self-supervised learning on images or point clouds are increasingly used as a general-purpose interface for vision models. The connection between the two has been explored for subject specific editing, aligning and training video diffusion models, but not in the role of a dense conditioning signal for pretrained video diffusion models. Features obtained through self-supervised learning like DINOv3, contain a lot of entangled information about style, lighting and semantics of the scene. This makes them great at reconstruction tasks but limits their generative capabilities. In this paper, we show how we can use the features for tasks such as video domain transfer and video-from-3D generation. We introduce a lightweight control architecture and training strategy that decouples appearance from other features that we wish to preserve, enabling robust control for appearance changes such as stylization and relighting. Furthermore, we show that low spatial resolution can be compensated by higher feature dimensionality, improving controllability in generative rendering from explicit spatial representations.
arXiv:2607.01235v1 Announce Type: new Abstract: Understanding how Large Language Models (LLMs) make token-level decisions during code generation remains a major challenge for both researchers and practitioners. While recent tools provide insights into model internals or generation outcomes, they often lack decoding-time signals, fine-grained uncertainty measures, and interactive mechanisms for exploring alternative generation paths. We present TokenScope, an interactive interpretability and analysis tool for decoder-based LLMs that exposes token-level metrics, attention patterns, and structural information during generation. TokenScope supports interactive token replacement, counterfactual branching, and code-aware aggregation via abstract syntax trees. By unifying decoding-time signals with structural program analysis, TokenScope enables systematic investigation of LLM behaviour during code generation.
arXiv:2607.01236v1 Announce Type: new Abstract: As LLM agents gain increasing access to powerful tools, ensuring that their actions are aligned with the user's intent becomes critical. When an agent's proposed tool invocation deviates from the user's intent -- a phenomenon called misalignment -- it may lead to harmful consequences that are difficult to undo. Existing runtime guardrails rely on an LLM-as-a-judge paradigm that lacks a systematic framework for reasoning about alignment, often producing judgments that are inconsistent or difficult to audit. Motivated by provenance analysis, we propose a provenance-based conceptual framework that formalizes misalignment detection as determining whether a proposed tool call is supported by traceable evidence in the agent's context. Building on this framework, we propose ProvenanceGuard, a multi-stage pipeline that analyzes the agent's action for three types of misalignment before the selected tool is executed and only allows the action to take place when it is considered aligned with the user's input query. We evaluated our proposed approach on two different benchmarks, Agent-SafetyBench and WorkBench, across 10 backbone LLMs. Compared to the LLM-as-a-judge baseline, ProvenanceGuard reduces error rate on misaligned traces from 42.9% to 1.8% on Agent-SafetyBench and from 32.1% to 17.3% on WorkBench, while reducing intervention burden on task-successful traces from 30.5% to 12.8% and introducing no statistically significant increase in unnecessary interventions on aligned traces. These results demonstrate that structured, provenance-based reasoning provides an effective and practical foundation for safeguarding LLM agents from misalignment.
arXiv:2604.02444v2 Announce Type: replace Abstract: Real-world table question answering often involves hybrid schemas in which some query-relevant information is explicit in relational columns, while other attributes, predicates, or join conditions are only implicit in free-form text. Existing systems struggle with this setting: Text-to-SQL methods scale to large and multi-table databases but require fully structured schemas, whereas direct LLM-based methods can interpret textual content but are costly and unreliable when applied to large databases. We present OmniTQA, a unified framework for semi-structured table question answering that treats semantic reasoning as a first-class operation within relational query execution. OmniTQA compiles natural-language questions into directed acyclic graphs of relational and LLM-based semantic operators. This enables ambiguity-aware plan diversification, cost-aware optimization, and dual-engine execution over structured and textual data. Across structured and semi-structured benchmarks, OmniTQA consistently improves performance in hybrid settings, outperforming the strongest baselines by 14 accuracy points on average and by 27 points on the most challenging subset, while maintaining competitive accuracy on fully structured datasets.
arXiv:2602.18803v2 Announce Type: replace Abstract: We present LoTIS, a model for visual navigation that provides robot-agnostic image-space guidance by localizing a reference RGB trajectory in the robot's current view, without requiring camera calibration, poses, or robot-specific training. Instead of predicting actions tied to specific robots, we predict the image-space coordinates of the reference trajectory as they would appear in the robot's current view. This creates robot-agnostic visual guidance that easily integrates with local planning. Consequently, our model's predictions provide guidance zero-shot across diverse embodiments. By decoupling perception from action and learning to localize trajectory points rather than imitate behavioral priors, we enable a cross-trajectory training strategy for robustness to viewpoint and camera changes. We outperform state-of-the-art methods by 20-50 percentage points in success rate on conventional forward navigation, achieving 94-98% success rate across diverse sim and real environments. Furthermore, we achieve over 5x improvements on challenging tasks where baselines fail, such as backward traversal. The system is straightforward to use: we show how even a video from a phone camera directly enables different robots to navigate to any point on the trajectory. Videos, demo, and code are available at https://finnbusch.com/lotis.
arXiv:2602.20358v2 Announce Type: replace Abstract: The study of stable matchings usually relies on the assumption that agents' preferences over the opposite side are complete and known. In many real markets, however, preferences might be uncertain and revealed only through costly interactions such as interviews. We show how to reach interim-stable matchings, under which all matched pairs must have interviewed and agents use expected utilities whenever true values remain unknown, while minimizing both the expected number of interviews and the expected number of interview rounds. We introduce two adaptive algorithms that produce interim-stable matchings: one operates sequentially, and another is a hybrid algorithm that begins by scheduling some interviews in parallel and continues sequentially. Focusing on cases where agents are ex-ante indifferent between agents on the other side, we show that the sequential algorithm performs 2 interviews per agent in expectation. We complement this by showing that any algorithm that performs less than 2 interviews per agent, does not always guarantee interim-stability. We also demonstrate that the hybrid algorithm requires only polylogarithmic expected number of rounds, while still performing only about 2 interviews per agent in expectation. Additionally, the interviews scheduled by our algorithms guarantee an interim-stable matching when Deferred-Acceptance is run after all interviews are completed.
arXiv:2607.01243v1 Announce Type: new Abstract: The rapid growth of scientific publications makes it increasingly difficult to track and synthesize research progress. While Large Language Models (LLMs) can support automated survey generation, existing methods retrieve unstructured data and require models to infer conceptual, methodological, and taxonomic relations from raw text at generation time. We introduce STRUCTSURVEY, a hierarchical multi-agent framework that shifts structural reasoning from generation to retrieval by dynamically constructing graph-based representations of entities, relations, and topical taxonomies. We evaluate STRUCTSURVEY on a new reference-grounded benchmark of ACL survey papers for reproducible long-form scientific summarization. Compared with embedding-only retrieval baselines, STRUCTSURVEY improves ROUGE-1 recall by +2.9 and ROUGE-2 recall by +1.0 on average, without reducing precision. It also improves LLM-as-a-Judge ratings for logical structure, depth, and synthesis, showing that explicit structural retrieval yields surveys closer to human-written organization and reasoning.
arXiv:2607.02364v1 Announce Type: new Abstract: We study the problem of deterministically computing the exact root of a sparse polynomial in the multivariate setting. Let $f \in \F[x_1,\ldots,x_n]$ be a nonzero polynomial that is an exact $e$-th power, say $f = g^e$. Suppose $f$ is $s$-sparse, has an individual degree of at most $d$, and a total degree of $D = \tdeg(f)$. We prove a sparsity bound on the base polynomial $g$: \[ \|g\|_0 \le s^{D(2d+2)/e + 1}. \] Based on this bound, we develop a deterministic algorithm that computes the base $g$. % In contrast to the general deterministic factorization algorithm of Bhargava, Saraf, and Volkovich \cite{BhargavaSarafVolkovich2020}, which achieves only a quasi-polynomial dependence on the input parameters, our algorithm is \emph{polynomial-time} in the setting where the total degree $D$ is bounded. Specifically, the overall complexity is \[ \mathrm{poly}\left(s^{O(Dd)}, n, d, D\right) + s\cdot R(e), \] % where $R(e)$ denotes the cost of constructing a single $e$-th root of a scalar in the base field $\F$, and, when $\operatorname{char}(\F)\mid e$, the cost of computing a single Frobenius root of a scalar. % This term is field-dependent, and over finite fields, $\mathbb{Q}$, or number fields with a suitable representation, it is absorbed into the polynomial complexity bound. % Within the bounded total-degree regime, this yields a deterministic polynomial-time algorithm for exact-root computation.
arXiv:2607.01252v1 Announce Type: new Abstract: Background: Dermatology AI has mainly focused on image-based diagnosis, while chronic disease workflows have received less attention. We surveyed Indian dermatologists to map routine clinical challenges, with a focus on atopic dermatitis (AD), and assess current AI use. Methods: A nationwide cross-sectional survey commissioned by the Society for Eczema Studies included 377 practicing Indian dermatologists. The survey assessed clinical challenges, AD workflow barriers, AI use, adoption barriers, and ethical concerns. Analyses used descriptive statistics, chi-square tests, false discovery rate correction, and multivariable logistic regression. Results: Patient adherence (61.3%) and treatment planning in difficult or refractory cases (57.0%) were reported more often than diagnostic uncertainty (48.0%). In AD care, severity scoring was reported as a challenge by 47.7% and had the lowest satisfaction among measured workflow areas. Current AI use was reported by 49.9%, most often involving general large language models for literature synthesis, documentation, and academic tasks rather than specialized image analysis. Barriers differed by experience: dermatologists with more than 20 years of practice more often cited lack of training, while those with 5 years or less more often cited lack of clinical utility after trying AI tools. AI users were more likely than non-users to report concern about patient self-misdiagnosis and anxiety, which remained significant after adjustment for experience and academic affiliation. Conclusion: Respondents reported using general-purpose AI mainly for cognitive and administrative tasks, while their clinical needs centered on chronic disease management and AD workflow support. Clinician-supervised workflow tools may be more useful than standalone diagnostic applications.
arXiv:2607.01283v1 Announce Type: new Abstract: Grid-based approaches to approximate nearest neighbor (ANN) search have been absent from modern scaling analyses. We present a systematic characterization of a multiprobe grid algorithm with respect to dataset size $N$ and dimensionality $d$. Our experiments reveal a previously unreported $d$-scaling crossover on the GloVe embedding family, in which multiprobe grid search maintains an approximately constant dimensional scaling exponent while other graph-, tree-, and partitioning-based methods exhibit degrading throughput. The advantage comes with near-linear query scaling in $N$, but also with lower indexing cost than competing ANN methods. Our results suggest that grid-based methods such as multiprobe grid may be competitive in rebuild-heavy or high-dimensional settings where indexing cost and dimensional robustness dictate performance. More broadly, recent work has formalized self-attention as an ANN operation. Thus, the $N$- and $d$-scaling properties of ANN algorithms may guide cost analysis of efficient transformer architectures. Code is available at: https://github.com/weiz345/MultiProbeANN.
arXiv:2607.02437v1 Announce Type: new Abstract: Time series forecasting remains challenging when the underlying data contain rare but critical extreme events. This issue is particularly important in hydrologic forecasting, where streamflow distributions are often highly skewed and extreme peaks can have substantial impacts on flood monitoring, water resource management, and early warning systems. Although Transformer-based forecasting models have achieved strong performance by modeling long-range temporal dependencies, they typically treat all time points uniformly and may therefore underrepresent rare extreme patterns. In this paper, we propose the Extreme-Adaptive Transformer (Exformer), a forecasting framework designed to explicitly model temporal dependencies involving both normal and extreme events. Exformer introduces an extreme-adaptive attention mechanism composed of three sparse components: Local, Stride, and Extreme. The Local and Stride components capture short-term and periodic temporal dependencies, respectively, while the Extreme component selectively models event-aware dependencies between normal and extreme streamflow patterns. Experiments on four real-world hydrologic streamflow datasets show that Exformer achieves superior 3-day forecasting performance compared with state-of-the-art baselines. Our findings demonstrate that explicitly incorporating extreme-aware attention improves the forecasting capacity of Transformer models on imbalanced time series with rare but consequential events.