Papers

MeanFlow predicts interval-average velocities, but conventional few-step sampling ties the interval queried from the model to the duration of the state update. We separate these two quantities at inference, fixing the model, update grid, and number of queries while varying the prediction horizon. Our central finding is that the horizon response depends on both query position and the states entering that query. Across seven independently trained CIFAR-10 checkpoints at classifier-free guidance (CFG) scale 2, a longer horizon worsens FID at early steps but improves it at the midpoint. To probe whether query time alone determines this response, we fix the midpoint query and current/subsequent guidance while changing only earlier CFG. In held-out factorial outcomes on five previously studied checkpoints, the mean midpoint horizon benefit increases from 0.109 to 1.308 FID, indicating sensitivity to the incoming-state distribution. A frozen four-query rule reduces mean CIFAR-10 FID by 2.023 without retraining or extra queries, including gains on five checkpoints trained after rule selection. A milder rule also improves mean FID by 1.041 on a stronger frozen grid, with fixed-rule FID gains on the evaluated public ImageNet-256 and AFHQ checkpoints. However, matched-budget calibration favors query-start shifting in historical CIFAR tests, and the tested Horizon allocation adds no FID benefit beyond Grid+Start on two public ImageNet models. Thus, prediction horizon is a distinct, position- and state-sensitive inference control, but not a universally preferred one.

Internet Exchange Points (IXPs) operate from colocation facilities that draw power from regional grids and store limited backup fuel. In April 2025, the Iberian blackout kept the Spanish grid down for 16 hours, beyond the 12 hours of fuel required by the industry standard, while traffic at major Iberian IXPs fell by up to 70%. Yet operators cannot tell which facilities a grid event cuts off, whether it outlasts their backup, or where traffic from a failed IXP goes. We present SPARC, which maps power-grid and Internet dependencies from substations to IXPs and their traffic. Using open data validated against transmission operators, SPARC links 552 facilities hosting 275 IXPs to the substations that likely feed them. We spread grid events over this map and measure failover through controlled BGP withdrawals on the PEERING testbed. Disabling the most-shared 16% of facility-feeding substations cuts power to half of all IXP presences (an IXP at one facility), under half of what randomized maps require. Losing the 5% most central substations cuts power to 0.1% of IXP presences directly but to 9-52% once the grid splits into disconnected islands. Traffic often fails over to nearby sites that share the failing grid: a cascade affecting four metropolitan areas sends up to 40% of one area's monitored addresses to another of the four. We introduce SPARC-Shield to select substations to harden or facilities whose backup to extend. Against an attacker disabling ten substations, hardening the ten selected by SPARC-Shield reduces offline IXP presence by 72%, compared with 52% for the best standard ranking.

What makes a deep network train, and what shape is its loss surface? We answer both with one construction at two scales: a polynomial activation f (z) = z + {\lambda}z 3 , clamped and projected onto the unit sphere, plus an unnormalized identity path. As a deep stack it trains to one hundred layers on MNIST with no normalization machinery; as a small model its square loss is a polynomial system we solve exactly. The small scale explains the large one: it prescribes how to design, initialize and train the stack.

LLM agents increasingly design a system before implementing it, yet the data to train and evaluate them centres on implementation and testing: requirements and design artefacts are often missing, outdated or private. Source code, in contrast, is abundant. Prior work recovers architecture documentation by feeding LLMs the source code or letting agents explore it, which saturates the context, does not scale to many repositories, and yields a single artefact rather than aligned ones. We present a framework that recovers aligned stakeholder intent, requirements and architecture from a compact repository representation, a static dependency graph and key contextual files, and release the artefacts of 10,764 Java, Python, C/C++ and C# repositories as a dataset. An LLM judge rates the artefacts of three generators as clear and internally consistent, but less reliably correct and at the right level of abstraction. To test whether the artefacts carry the design, we pose a software Ship of Theseus: a coding agent rebuilds 29 Commit0 libraries from the intent alone or with the recovered requirements and architecture, given only an interface contract and part of the tests. The artefacts do not significantly change the held-out pass rate (33.7% vs. 35.8%), but bring the rebuilt module structure significantly closer to the original (module F1 from 0.54 to 0.63), with either generator and even with concise artefacts. Recovered artefacts thus convey design that tests do not capture, offering a scalable source of design-level data for agents that design before they implement, and the round trip measures this without human references.

Large Language Models (LLMs) frequently encounter conflicts between internal parametric knowledge and external contextual knowledge, yet the mechanisms governing this selection remain unclear. We show that the choice between context and parametric knowledge is linearly represented and can be controlled along a single direction in the model's residual stream. We first demonstrate that this direction generalizes across factual-conflict settings, including context formats, knowledge categories, and datasets, while consistently shifting behavior between parametric and contextual answers. We then transfer a direction learned from simple factual conflicts to three more complex task families: RAG, memory, and rule following, where models must acquire information from retrieved documents, interaction histories, or explicit rules. Across four model families and multiple transfer settings, steering consistently improves the use of prompt-provided information over the unsteered model. These results suggest that the direction captures a transferable preference for external knowledge rather than a mechanism specific to factual conflicts.

In a repeated Stackelberg security game, a defender commits in every round to a randomized allocation of its resources, and an attacker, drawn from a set of types, best-responds. Existing no-regret guarantees assume that the defender knows this set of types in advance. We drop this assumption: the set of types is unknown, and its size is at most $K_{\max}$, an upper bound that the defender is not told. We show that what can be learned then depends on what the defender observes. If it observes the type of each attacker, an algorithm that adds types to its list as they appear achieves regret $\tilde O(nK_{\max}\sqrt{T})$ over $T$ rounds with $n$ targets, the same $\sqrt T$ rate as with known types, at an extra factor of only $\sqrt{K_{\max}}$ and without ever using $K_{\max}$. If it observes only the attacked target and types come from an infinite universe, no algorithm has sublinear regret: in a fixed three-target game every algorithm suffers regret at least $T-2$, even against a single attacker type, because it cannot tell which type it faces. If, in addition, it is told every $B$ rounds which types have appeared so far, sublinear regret returns, at $\tilde O(nK_{\max}^{4/3}T^{2/3}+K_{\max} B)$. The first term is the rate of the partial-information learner we use, and we show that the additive cost of the delay cannot be avoided.

We present a two-year homogeneous optical monitoring campaign of three TeV-emitting high-synchrotron-peaked BL Lac objects, 1ES 1959+650, H 1426+428, and PG 1553+113, carried out with the Asiago Schmidt 67/92 cm telescope. TeV blazars are key targets in high-energy astrophysics, relevant for relativistic jet physics, particle acceleration, multiwavelength variability studies, and, in selected cases, multimessenger programmes. In this context, dense optical monitoring provides a low-energy synchrotron tracer of jet activity over temporal baselines that are difficult to cover continuously with high-energy facilities. The dataset consists of more than 17,000 R- and I-band photometric measurements obtained over 957 source-filter observing sessions between 2024 February 1 and 2026 January 31, for a total monitoring time of about 420 h. We applied a uniform differential-photometry and classical F-test framework based on two independent comparison-star pairs. Robust IDV is detected only in H 1426+428, with 15 out of 43 analysed sessions classified as variable in the R band and 5 out of 40 in the I band. No significant IDV is detected in 1ES 1959+650 or PG 1553+113 within the adopted statistical framework. For H 1426+428, the corresponding duration-weighted duty cycles, computed using only variable sessions with a measured local characteristic timescale, are 28.0% in the R band and 7.0% in the I band. The inferred local extrapolated doubling/halving timescales are of the order of 5-15 min in the observer frame. These results indicate that rapid optical synchrotron variability in TeV HBLs is intermittent and state-dependent. They also show that robotic small- and medium-aperture optical facilities can provide valuable long-term support for high-energy monitoring programmes by supplying dense, uniform, and repeatable optical coverage.

We present FoveaVLA, a Vision-Language-Action (VLA) model that replaces the wrist camera with learned foveation actions. Wrist cameras provide close-up detail but restrict the view to the end-effector's surroundings. External cameras see the whole scene, but the policy receives them at low resolution, so the fine detail needed for manipulation is lost. FoveaVLA recovers this detail by predicting, alongside each action chunk, the interaction region to observe at the next policy query. The policy then receives crops of these regions, taken from the original-resolution images, together with the full views. We build FoveaVLA on $\pi_{0.5}$ in two configurations, Single and Dual, using one or two external stereo cameras with full and foveal RGB and depth inputs, and post-train both on DROID with RoboInter interaction annotations. Across four real-world task suites, FoveaVLA-Single achieves 70.8% success with one external RGB-D view, and FoveaVLA-Dual achieves 77.1%, versus 52.1% for the wrist-equipped $\pi_{0.5}$. Both match the baseline on plain pick-and-place and improve success with distractors, cloth manipulation, and far-away objects.

Defect engineering is a crucial tool for tuning and tailoring the properties of two-dimensional (2D) materials. A key challenge is identifying which defects are kinetically stable at room temperature and sufficiently promising to warrant further investigation. However, identifying stable defects with suitable properties using first-principles calculations, such as density functional theory (DFT) is computationally demanding, effectively limiting the scope of high-throughput studies. In this work, we investigate whether the machine learning interatomic potential MACE can reproduce sorption trends (adsorption vs. absorption) previously obtained from high-throughput DFT calculations and extend the analysis to elevated temperatures. A total of 17,624 defect structures from the ADAQ database are simulated using MACE-accelerated molecular dynamics (MD) at both 0 K and 300 K. The results show that MACE successfully reproduces the overall sorption trends observed in DFT calculations at 0 K at a fraction of the computational cost. Simulations performed at 300 K further indicate that the qualitative sorption trends remain largely the same when including finite-temperature effects. These findings demonstrate that MACE-driven MD can be used as an effective high-throughput prescreening tool for defect-related studies in 2D materials, specially related to sorption. While conventional first-principles methods, such as DFT, remain necessary for accurate and quantitative validation, the use of MLIPs enables MD simulations of finite-temperature defect behavior at a fraction of the computational cost of DFT-based MD.

Computing the top-$k$ simple shortest paths between two vertices of a graph, i.e., solving the P-KSSP problem, is a fundamental primitive in graph analytics. In modern scenarios, however, queries are rarely one-shot: parameters change as applications adapt their information needs, and recomputing a solution from scratch at every request quickly becomes a bottleneck. We study P-KSSP in a dynamic setting, focusing on changes of the parameter $k$: given a top-$k$ solution for a pair of vertices and a new value $k'$, the goal is to compute a top-$k'$ solution reusing the previously computed paths. Our approach rests on the pseudo-tree, a compact representation of a collection of simple $s$–$t$ paths: we show that building it from an arbitrary, unstructured solution amounts to trie construction, and give two constructions running in time linear in the size of the collection. On top of it, we design UPD, a dynamic algorithm that updates a solution upon a change of $k$; we prove its correctness, analyze its complexity, and show that it is a general, solver-agnostic scheme, which starts from the paths alone, is parametric in the routine that generates candidate paths and combinable with known preprocessing techniques, as well as with an online pruning rule that we formalize. While a worst-case update is asymptotically comparable to recomputing from scratch, an extensive experimental evaluation shows that a variant of UPD is the fastest method in the large majority of tested configurations, often by a wide margin (with speedups over recomputation that grow with $k$ and $k'$ and exceed $60\times$), yielding a practical recipe for the configuration to deploy.

Photon chirality is inherently sensitive to structural handedness and magnetism, which break spatial inversion and time-reversal symmetries, respectively. Ferroaxial order, by contrast, hosts two-dimensional chirality while preserving both symmetries, thereby precluding its direct coupling to photon chirality. Intertwining ferroaxiality with magnetic order, however, may lift such constraints and enable unconventional coupling to photon chirality. Nevertheless, the optical response arising from such intertwined order has yet to be demonstrated experimentally. Here, we demonstrate transverse magneto-gyrotropy arising from spin-driven ferroaxiality in the Weyl magnet EuCd2Sb2. Whereas the conventional polar magneto-optical Kerr effect is governed by magnetization parallel to the light-propagation direction, we observe a gyrotropic response associated with magnetization transverse to the light-propagation direction. The optical rotation exhibits a characteristic 120-degree periodicity upon rotation of the in-plane magnetic field, reflecting the threefold symmetry of the emergent ferroaxial state and revealing the ferroaxial origin of the transverse gyrotropic response. This coupling to photon chirality enables vectorial control of the gyrotropic response through the magnetic field direction. Our findings establish an interaction between photon chirality and transverse magnetization enabled by intertwined ferroaxial and magnetic orders, demonstrating a novel form of multiferroicity with functional light-matter interaction.

We consider the big finitistic projective and flat dimensions of a commutative ring and of its formal power series ring. When the ring is Noetherian, both invariants increase by one. Coherence of the power series ring does not force either equality to hold. For every \(n\ge0\) and every \(d\) with \(n+1\le d\le\infty\), the flat pair \((n,d)\) can be realized. The finite projective pairs are realized when \(n\ge1\), and also for \((0,1)\). For every \(n\ge0\), one family realizes \((n,\infty)\) for both invariants at the same time. The seed is a zero-dimensional perfect local ring. Its power series ring has countably presented modules whose flat dimension is \(r\) and whose projective dimension lies between \(r\) and \(r+1\), for every \(r\ge1\). The remaining finite projective case is controlled by a dichotomy for perfect coefficient rings. If the Jacobson radical is nilpotent, the power series value is one; if it is not nilpotent, there is a module of projective dimension exactly three. In particular, the pair \((0,2)\) is impossible. For one fixed countable perfect ring, the power series ring has finitistic flat dimension one. Its finitistic projective dimension is \(m+2\) when \(2^{\aleph_0}=\aleph_m\) with \(1\le m<\infty\), and is infinite when \(2^{\aleph_0}>\aleph_\omega\).

Context: Circumstellar gas around young stars provides key insights into planet formation and disk evolution. HD 152384, a young (5 Myr) A0 V star in Sco-Cen, shows hybrid-disk traits: a modest infrared excess from small dust grains and atomic emission lines dominated by refractory elements, but its disk structure and origin remain unclear. Aims: We characterise the kinematics and excitation of the circumstellar gas around HD 152384 and constrain the physical properties of its emitting region. Methods: We use new high-resolution optical and near-infrared VLT/UVES and VLT/CRIRES spectra to re-examine the circumstellar environment of HD 152384, previously suggested to host a broad, gas-rich disk. Results: The observations reveal 74 narrow, double-peaked emission lines of Fe, Mg, Si, and Ca, with similar profiles and a 103 km/s peak separation. Keplerian modelling places the gas in a geometrically thin ring at 0.09-0.12 au, close to the dust sublimation radius (0.10-0.14 au for a sublimation temperature of about 1500 K). The Fe I emission is optically thick, whereas the optically thin Mg I lines imply 2.2 x 10^22 kg (3.6 x 10^-3 Earth masses) of magnesium gas in the ring. Conclusions: The dust and gas properties are consistent with a secondary origin involving collisions of rocky planetesimals. We propose that small collision-produced dust grains spiral inward and sublimate near the dust sublimation radius, accumulating atomic gas there. HD 152384 thus represents a young analogue of gas-bearing debris disks such as Beta Pictoris and gaseous debris disks around metal-polluted white dwarfs, offering a new window on the early violent processing of rocky planetesimals.

OpenAI (using a non-public "internal frontier model”) has recently (6 October 2026) made a number of remarkable claims, including some claims regarding the quasi-Riemann hypothesis (QRH). While there are good reasons to be cautious regarding these claims, one can nevertheless ask what (if true) their implications would be. While there is widespread agreement that any version of the QRH will provide various improved bounds on the location of the prime numbers, I feel that there is some value in making these improvements explicit. Specifically, if we assume validity of the claimed 7/8 version of the QRH then I shall demonstrate (for all $n\in \mathbb{Z^+}$): 1) the existence of primes between eighth powers, in the interval $(n^8,(n+1)^8)$; 2) the existence of primes in the shorter interval $(n^8,n^8+8n^7)$; 3) the existence of primes in the even shorter interval $(n^8,n^8+n^7).$ Similarly for reals greater than unity $x>1$ one can demonstrate: 4) the existence of primes in the Hoheisel-style interval $(x,x+x^{7/8})$. All of this is contingent on the validity of the claimed 7/8 version of the QRH.

Reasoning language models increasingly rely on test-time compute scaling, yet most existing work studies how to allocate inference computation while treating the model itself as fixed. We ask a complementary question: can reasoning models be post-trained to make better use of a given amount of test-time compute, and which examples are most useful for this goal? Our key observation is that reasoning problems differ substantially in their response to additional computation. Some are already solved under limited compute, others remain unresolved even after further reasoning, while a distinct subset changes from complete initial failure to success when additional reasoning is provided. We call these compute-responsive (CR) examples and introduce Compute-Responsive Data Selection (CRDS), a post-training data-selection principle that prioritizes examples with empirically demonstrated marginal utility from additional reasoning. On Qwen3-8B, SFT on CR examples recovers 45.3% of held-out low-compute failures, compared with 40.6% when training on already-solved examples. The advantage extends beyond the selected CR split: SFT-CR reaches 51.8% overall accuracy on an independently sampled test distribution, with the gain concentrated on compute-responsive problems. CR training also shifts the compute–accuracy frontier across inference budgets from 512 to 4,096 tokens, outperforming matched non-CR training throughout and reaching 48.1% recovery at the largest budget. The same pattern is supported across multiple data sources, remains visible under on-policy distillation, and extends to Llama-3.1-8B-Instruct when CR sets are reconstructed separately for that model. Better test-time compute utilization can therefore be learned through post-training, with recoverable responses to additional reasoning providing a particularly informative signal for selecting training examples.

Solar flares serve as standard candles for all magnetic-reconnection-based energy releases, and their consequences into the heliosphere can be modeled from first principles. This review presents the latest developments in modeling solar flares and their associated flux rope expulsions with state-of-the-art multidimensional magnetohydrodynamic (MHD) descriptions. Observationally established ingredients beyond a single-fluid, thermal plasma evolution have been incorporated to varying degrees, and we highlight self-consistent efforts realized in "standard flare" setups.

The rapid advancement of large vision language models (LVLMs) has made geolocation inference from images highly accurate and accessible. This poses serious privacy risks: a passive adversary with black-box LVLM access can infer a user's location from a single photo with no metadata. Existing protection methods rely almost exclusively on adversarial perturbation, which introduces perceptible visual artifacts and utility degradation. We identify the visual-geolocation asymmetry: geolocation-decisive clues, such as street signs, tend to be visually peripheral but dominate location inference. Exploiting this asymmetry, we propose GeoCloak, an agentic framework for semantics-preserving geolocation privacy protection. It acts as a clue-aware planner that analyzes each image to identify geolocation-relevant clues, dispatches targeted semantic edits to directly suppress peripheral clues, and autonomously configures adversarial perturbation to further suppress residual geolocation features, all under iterative utility-aware feedback. We evaluate GeoCloak on four diverse datasets spanning 108 countries and 576 cities against five LVLMs and three inference methods. GeoCloak reduces overall city-level accuracy from 41.2% to 9.9%, approximately half of the best-performing baseline, while achieving substantially superior utility preservation both qualitatively and quantitatively. We hope this work takes a meaningful step toward practical, semantics-preserving geolocation privacy protection, giving users greater control over the location information inadvertently embedded in their shared images.

The horizontal mean curvature flow is a geometric evolution associated with the horizontal perimeter in sub-Riemannian spaces. In this paper, we introduce a minimizing movement scheme for this flow in the first Heisenberg group $\mathbb{H}$, and prove the existence of a flat flow which is locally $1/2$-Hölder continuous in time, with respect to the $L^1$-distance. The construction applies to arbitrary bounded sets of finite horizontal perimeter and does not require smoothness or symmetry of the initial datum.

Regression models used in real-world applications are often challenged by noise, outliers, data heterogeneity, and asymmetric error characteristics, which severely limit the effectiveness of conventional point-wise learning models. While twin support vector regression (TSVR) partially addresses these issues through the use of nonparallel regression bounds, it still relies on individual samples as basic learning units. As a result, TSVR remains highly sensitive to noisy and contaminated observations and fails to leverage the underlying structural and regional information inherent in complex datasets, leading to suboptimal robustness and generalization performance. To address these limitations, we introduce granular ball twin support vector regression (GBTSVR) and its quantile extension, granular ball twin support vector quantile regression (GB-TSVQR). The proposed models integrate granular ball computing with TSVR, replacing individual data points with compact, structure-preserving granular balls as fundamental learning units. This region-level representation effectively suppresses noise and outliers while preserving intrinsic data geometry, thereby enhancing robustness, stability, and generalization. GBTSVR constructs nonparallel lower- and upper-bound regression functions over granular balls, significantly reducing sensitivity to sample-level perturbations. Building upon this framework, GB-TSVQR further incorporates quantile regression principles to simultaneously capture heterogeneous trends and asymmetric error distributions across different quantile levels. Extensive experiments on synthetic datasets, time-series benchmarks, and outlier-contaminated data demonstrate that the proposed GBTSVR and GB-TSVQR consistently outperform state-of-the-art regression models.

Ambiguous inputs can admit multiple interpretations, each potentially yielding a different answer. Existing work largely treats this ambiguity as a property of the input, yet for a particular language model, only some alternatives may be known and answerable. We therefore distinguish world ambiguity, capturing all alternatives admitted by an input, from model ambiguity, capturing those accessible to the responding model. To characterize these alternatives, we formalize natural-language clarifications through three desiderata: uniqueness, faithfulness, and answer equivalence. We show that these desiderata induce a one-to-one correspondence between valid answers and equivalence classes of clarifications, implying that any interpretation can be represented either by its answer or by the clarifications that uniquely express it. We leverage this correspondence to introduce \circleclar{}, a black-box method that estimates the set of answers available to a model by checking consistency along an answer–clarification–answer cycle. Across open-domain QA and executable Text-to-SQL, \circleclar{} achieves strong recovery of these answer sets. In QA, the recovered support yields the strongest overall performance in model-dependent ambiguity detection and enables effective selection among answering, clarifying, and abstaining. Together, these results demonstrate the practical value of a model-dependent view of ambiguity and its characterization through an answer-clarification correspondence.

In recent years, with advances in multimodal interaction technologies and the growing demand for intelligent emotional support, Multimodal Emotional Support Conversation (MESC) has emerged as an important task. However, existing systems still suffer from shallow responses and inappropriate support strategies due to weak semantic feedback, cross-dimension reward compensation, and coarse reasoning supervision. To address these challenges, we propose FACES, a fine-grained gated semantic alignment framework for MESC. Specifically, FACES integrates multi-source rewards to construct multidimensional semantic and functional alignment signals, employs a non-compensatory gated fusion mechanism to reduce cross-dimension reward offset, and leverages structured Chain-of-Thought (CoT) anchors to perform token-level credit assignment. Under a controlled response-length protocol, FACES improves the average MLLM-based evaluation score from 3.61 to 4.04 and achieves an 80.22% pairwise win rate over SOTA. Human evaluation under the same protocol yields a 75.3% preference rate. Results on three external datasets further support transferability. The code and dataset are available at https://github.com/OCT-ARR/arr.

We study equilibrium fluctuations for the symmetric partial exclusion process (SPEP) on the one-dimensional integer lattice $\mathbb{Z}$, an interacting particle system in which each site is occupied by at most $\alpha$ particles. We assume that the jump rate is decomposable and that the instantaneous current admits a gradient form. Our aim is to study the scaling limit of the density fluctuation field of the SPEP starting from an invariant measure and to derive the corresponding stochastic partial differential equation (SPDE). A distinctive feature of our setting is that the maximum number of particles per site depends on the scaling parameter $n$ as $\alpha=\alpha_n \propto n^a$ for some $a>0$, so that the model interpolates between the partial exclusion process ($\alpha <\infty$) and the symmetric zero range process (SZRP, $\alpha = \infty$). Under the diffusive rescaling in which the lattice spacing is $1/n$ and time is accelerated by a factor $n^2$, we prove that the density fluctuation field converges to the solution of the stochastic heat equation. The limiting SPDE coincides with the one obtained for the SZRP. A key ingredient in the proof is the convergence, under a suitable rescaling, of the partition function of the reversible measure of the SPEP to that of the SZRP. This convergence implies that expectations of a wide class of local functions under the reversible measure of the SPEP converge to the corresponding expectations under the reversible measure of the SZRP, and provides a rigorous sense in which the SPEP approximates the SZRP as $n\to\infty$.

Agents extend their capabilities through reusable skills that package instructions, executable code, and resources. Composing these skills enables complex task workflows, but their dependencies can also connect malicious operations across skills, making their combined intent difficult to identify through individual-package auditing. We introduce CoRedSkill, a compositional red-teaming method that forms stealthy attacks through cooperating skills while preserving their original utility. CoRedSkill decomposes an existing workflow into interdependent skills and embeds attack operations in task functions through abstract syntax tree (AST) rewriting, with skill instructions guiding the agent to autonomously select and invoke the relevant skills, and workflow dependencies and intermediate artifacts connecting the injected operations during task execution. We construct a benchmark of 150 cases covering 50 tasks and three attack objectives to evaluate attack success, task completion, and audit detection. In the main evaluation, CoRedSkill achieves 98.0% attack success and 95.3% task success with a 3.3% detection rate, and a joint success rate of 90.7% (attack success, task completion, and audit evasion in the same case) compared with 6.7–42.7% for the four baselines. Skill compositions can therefore complete legitimate tasks while carrying out unauthorized operations that evade auditing.

Subject indexing assigns descriptors from a controlled vocabulary to bibliographic records, a task characterized by thousands of labels, few positive examples per label, and a long tail of rarely used descriptors. Existing extreme multi-label classifiers treat the vocabulary as a flat list, ignoring the semantic relations that thesauri encode. We present TAGHN, a heterogeneous graph reranker over a per-document subgraph of retrieved candidates augmented with the thesaurus neighborhood. The relevance of a candidate descriptor is estimated jointly from its similarity with the document and also the degree to which the candidate's thesaurus neighborhood matches the document content. A decomposed scoring head combines interpretable graph-derived features with retrieval signals. On STW, an economic thesaurus of 6,198 descriptors, TAGHN reaches 0.380 macro F1 against 0.346 for a fine-tuned cross-encoder and 0.278 for a fixed-vocabulary XMLC classifier. In terms of micro F1, it outperforms reranking models and stays competitive (-3.0) with the best fixed-vocabulary XMLC model. Qualitative analysis shows the graph resolving errors that per-label scoring cannot, selecting a narrower gold descriptor where an unstructured reranker settles for its broader parent.

As Internet of Things (IoT) devices collect and process data, security is essential. IoT devices often combine with other emerging technologies, such as artificial intelligence (AI) and fifth-generation (5G) communications, to support applications such as smart grids and internet-of-vehicles (IoV). Because IoT devices are deployed remotely, remote attestation (RA) effectively verifies each device's state. However, the scale of IoT deployments makes device management challenging for owners. Therefore, cloud-assisted IoT management services often provide RA solutions. Although cloud-based IoT services can securely manage IoT devices, passive malicious administrators or hacked IoT device owners' accounts can control IoT devices without cloud IoT service authorization. Therefore, this work proposes a cloud-assisted RA system that addresses three main challenges of IoT device RA: 1) lightweight RA design for resource-limited IoT devices, 2) unauthorized access to IoT devices by passive malicious administrators or compromised owner accounts on the cloud IoT management platform, and 3) RA may not extend to an authenticated key exchange protocol. In addition to proposing a swarm RA scheme that addresses these three challenges, this work analyzes the system's security properties and evaluates its feasibility through implementation and simulation.

We show that every vineyard can be decomposed into vines, that is, every path in the space of persistence diagrams $\mathcal{D}$ can be decomposed into a sequence of paths in $\mathbb R^2$. Moreover, we characterize which spaces $X$ have the property that every continuous map $X\to \mathcal{D}$ has such a decomposition. Our results hold not just for $\mathcal{D}$, but also for many other spaces of unordered tuples.

We study testing for peer effects in linear panel models with individual and time fixed effects when the interaction network is unspecified. We propose a quadratic-form test based on the Gram matrix of residualized peer covariates. A key difficulty is that fixed-effects removal and nuisance estimation destroy the centering of a conventional leave-diagonal-out quadratic form under conditional heteroskedasticity. We introduce a leverage adjustment that restores exact conditional centering and combine it with a residual-based variance estimator. The resulting studentized statistic is asymptotically standard normal as both panel dimensions diverge, without imposing any restriction on their relative growth rates. Under additional conditions, we establish consistency against alternatives. The null theory is further extended to covariates that are weakly dependent over time; the same studentization remains valid without an additional long-run variance estimator. Monte Carlo experiments evaluate the finite-sample size and power of the proposed test across different settings and demonstrate its favorable performance relative to the existing benchmark test. Applications to state tax competition and growth spillovers among OECD countries illustrate the practical relevance in testing for peer effects without specifying the interaction network.

To any compact holomorphically symplectic manifold and a nef BBF-isotropic class we associate a quasi-projective variety called the Markman–Markushevich space which parametrises the Lagrangian fibrations with this isotropic class modulo Tate–Shafarevich twists. We give a systematic interpretation of known results about Tate–Shafarevich twists of Lagrangian fibrations using this space, employ it to obtain a conjectural result for the last deformation type for which such a result is not yet proposed, and make some first steps towards its proof. As a byproduct, we generalise the results of Kumar on existence Shioda–Inose partners for abelian surfaces to a slightly more general class of K3 surfaces, and associate a canonical K3 surface with transcendental lattice isometric to $U^2\oplus\langle-2\rangle^2$ to a smooth cubic surface with a choice of a Schläfli double-six. We also review some other approaches towards the construction of complete families of O'Grady six-manifolds with Lagrangian fibrations, and show how they may still be useful even though they do not quite work.

We study the extended Dicke model obtained from the scalar–vector–tensor polarizability of open-shell atoms coupled to a shared cavity mode. For a single electric-dipole channel, rotational symmetry fixes the tensor to vector ratio; we use $^{168}\mathrm{Er}$ with $J=6$ as a product-state example. A circular pump produces a chiral Raman vertex whose tensor-dependent matrix elements select the occupied Zeeman sector. The resulting first-order transition occurs before the empty-cavity branch loses local stability, and the ordered state is read out through $\langle\hat J_z\rangle$. A linear pump instead produces an empty-cavity $\pm x$ cat and a continuous transition into an even-$m$ cat-like state with order parameter $\langle\hat J_z^2\rangle$. With cavity loss alone, the circular trajectory emits a photon burst and approaches a Raman-dark Zeeman pole rather than a finite-field condensate. In the linear geometry, cavity-bright modes are damped toward the cat-like branch while a dark spin mode remains undamped. These are phenomena of that model; the vector-only controls show that they depend on the channel-fixed tensor coupling.

We fix a spacelike two-dimensional left-invariant distribution and its associated almost contact structure on a three-dimensional Lorentzian Lie group, and investigate the star-Ricci contraction of the curvature of the Bott connection. We compute the complete tensor for seven standard bracket families and determine all real parameter solutions of the general algebraic equation and its self-adjoint subclass. Using the symmetric part of the tensor, we also define a Bott affine star-Ricci soliton equation and classify all left-invariant vector field solutions. Every affine solution is steady, while two non-unimodular families admit branches with a nonzero symmetrized tensor. The proofs provide complete component equations, treat all degenerate parameter branches, and establish sufficiency by explicit substitution. We also distinguish almost contact structures from contact metric structures.