Peer Revenue Now accepting positions

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

All of arXiv

Monday 9 November

Numerical Analysis

Convergence of finite element exterior calculus for the Hodge-Dirac and Hodge-Laplace problems on closed Riemannian manifolds

We prove that the finite element exterior calculus (FEEC) approximation to the Hodge-Dirac and Hodge-Laplace problems on any closed oriented Riemannian manifold converges. Its approximate solutions, together with their exterior derivatives, converge to the continuum solutions for every source, and at the rate of the mesh size when the source and its codifferential are square integrable. We use Licht's estimate for the FEEC approximation of forms with one square-integrable derivative by finite element forms. We combine this with a new abstract error bound for approximations of Hilbert complexes. This bound uses a version of Di Pietro and Droniou's third Strang lemma. To apply these estimates, we show that every closed manifold admits a family of arbitrarily fine triangulations whose shapes stay uniformly under control.
Read the full abstract

Distributed Computing

Optimal Proximity Black Hole Search with Programmable Matter

The concept of programmable matter (PM) envisions objects composed of tiny, mobile particles capable of self-arrangement, which allows them to change shape, color, and move. We initiate the study of the Black Hole Search problem within the context of PM. The problem involves a dangerous and unnoticeable location, the black hole, that, without a trace, destroys any particle visiting it. In our setting, particles move inside a convex area embedded in a triangular grid, where one vertex is a black hole. The objective is for the particles to identify a restricted set of vertices containing the black hole while ensuring that at least one particle survives. We present an algorithm utilizing three particles capable of assuming just two colors to identify the dangerous region where the black hole resides. Furthermore, we demonstrate that this algorithm is Pareto optimal in terms of number of particles and number of colors required, and asymptotically optimal in terms of number of epochs, where an epoch refers to the minimal time window within which each particle completed an activation cycle at least once.
Read the full abstract

Databases

Transactional Data Change for Real-Time Analytics in Lakehouses

Real-time data analytics has become a key demand for lakehouses. It requires propagating data changes from the data source to the lakehouse in real time, while ensuring ACID guarantees and minimizing resource consumption and impact on analytical query performance. This is highly challenging for a data system according to the RUM conjecture, especially given the read-optimized immutable file storage layer in lakehouses. Existing lakehouses offer only git-style catalog version control on a single table, and they trade data freshness, query performance, and/or resource efficiency to achieve a barely acceptable data-change throughput. In this paper, we propose a novel data synchronization framework, Retina, that enables real-time, cost-efficient, query-performant, and fully transactional data changes in lakehouses. We implement and open-source Retina. Evaluations show that Retina delivers 10-millisecond-level data freshness and over 3.2 million row/s linear-scalable data-change throughput with low resource consumption and query-performance overhead, significantly outperforming state-of-the-art lakehouse systems and lake-compatible HTAP databases.
Read the full abstract

All authors Dongyang Geng, Haoqiong Bian, Zinuo Li, Haoyue Li

Superconductivity

Quantum-geometry-controlled pseudogap and non-Fermi-liquid behavior

The quantum metric plays an essential role in flat-band superconductivity, governing the superfluid weight and coherence length. While previous studies have primarily focused on mean-field superconducting properties, the role of quantum geometry in fluctuation phenomena remains largely unexplored, despite the importance of fluctuations in flat-band systems. To address this issue, we employ a multiorbital T-matrix approximation that retains the orbital structure of the Bloch wave functions. By analyzing models in which the quantum metric can be tuned independently of the band dispersion, we show that an enhanced quantum metric suppresses the fluctuation-induced pseudogap in both flat-band and dispersive-band systems. In dispersive-band systems, this suppression is accompanied by a crossover from non-Fermi-liquid to Fermi-liquid behavior. This spectral evolution suggests a quantum-metric-driven crossover from a BEC-like strong-coupling regime to a BCS-like weak-coupling regime. More broadly, our results establish the quantum metric as a control parameter of interaction-driven fluctuation phenomena beyond mean-field theory.
Read the full abstract

Artificial Intelligence

FraudBench: A Legal Evaluation of AI Deception

We introduce FraudBench, an evaluation of large language models' (LLMs) propensities to recommend, assist, or directly take actions that would constitute fraud under U.S. law. Our evaluation contains a novel dataset of 159 scenarios based on real-world U.S. court cases and validated by legal experts. We present each scenario to LLMs via three tasks at varying levels of autonomy: advising users, drafting messages for users to send, and acting directly on behalf of users (477 tasks total). In each scenario, FraudBench re-enacts the facts of a court case to give LLMs an opportunity, but no explicit instruction, to either misrepresent or conceal a material fact in a high-stakes commercial setting–e.g., whether to describe company-funded product testing as "independent" to customers (misrepresentation) or whether to hide unfavorable regulatory investigations from clients (concealment). We evaluate 20 LLMs across 25 experimental conditions (totaling >300,000 generations) and include a preregistered human baseline ($n_{\texttt{annotations}} = 2{,}957$, $n_{\texttt{humans}} = 996$). We score a response as fraudulent when it engages in, assists with, or recommends misrepresentation or concealment. Key findings include: (1) across LLMs, fraud rates ranged from near 0% to over 40%; (2) LLM fraud rates increased with autonomy, with overall rates of 3.0% when advising, 14.7% when drafting, and 21.4% when acting directly, without applying pressure; (3) LLMs sometimes engaged in fraud despite recognizing legal risks; (4) LLMs more often engaged in fraud by concealment than by misrepresentation; and (5) prompt-based interventions substantially affected fraud rates, with a prompt to obey the law reducing fraud rates, while prompts to maximize profits and act agentically increased fraud rates.
Read the full abstract

All authors Kevin Wei, Sumaya Nur Adan, Stephan Llerena, Mark Gitau, Jakob Merane, Anita Srinivasan, Colette Le Brannan, Keethan Kleiner, Alessandro Tacconelli, Michael Tiu Jr., Case Thomason, Kyle Emili, Natasza Gadowska, Alex Mark, Eric Ren, Umang Bhatt, Jatinder Singh, Jacy Reese Anthis, Doni Bloomfield, Noam Kolt

Signal Processing

Joint Power and Adaptive-Constellation Optimization for OFDM-ISAC Systems

Next-generation wireless systems are expected to support integrated sensing and communications (ISAC). However, sensing imposes different requirements on the power allocation (PA) and the modulation format than communications, leading to a fundamental trade-off in frequency-selective channels. In this paper, we jointly optimize the PA and modulation format for orthogonal frequency division multiplexing (OFDM)-based ISAC systems, considering both target detection and tracking. We first derive an upper bound on the mutual information to predict the communication performance for power-constellation allocation. We then use an autoencoder (AE) to learn a family of finite constellations that realizes the allocation under different sensing constraints without retraining. The results show that tracking alleviates the sensing & communications (S&C) tradeoff compared with initial detection, and that the PA and the modulation dominate the S&C trade-off at low and high signal-to-noise ratios (SNRs), respectively. At intermediate SNR, the proposed joint optimization significantly outperforms individual optimizations, while the pretrained adaptive constellation family almost fully realizes the predicted allocation gains.
Read the full abstract

Materials Science

Misorientations in dislocation cell structures: Modelling and statistical analysis

Dislocation pattern formation is accompanied by the emergence of lattice rotations and concomitant misorientations across dislocation walls (cell boundaries). Recent advances in mesoscopic modelling of dislocation patterns enable for the first time a systematic simulation study of this phenomenon. We present simulation results for lattice rotation and misorientation patterns in crystals with face-centered cubic and rocksalt lattice structure. Statistical analysis of the orientation changes across cell walls reveals a transition associated with a symmetry-breaking bifurcation in slip system activity: Below the bifurcation, statistical tests indicate the presence of a single population of misoriented walls whereas, above the bifurcation, two sub-populations of 'incidental' and 'geometrically necessary' dislocation walls are present. Our data analytic approach demonstrates that this distinction does not require prior image segmentation and emphasizes the statistical robustness of the distinction between two types of dislocation walls.
Read the full abstract

Artificial Intelligence

Let the Fixer Decide: Training a Code Localizer from the Fixer's Likelihood

Code localizers are usually trained to predict the files a reference patch edits, but the code a fixer needs to read is not the same as the code it changes. We propose Fixer-Guided Localization (FGL): the localizer proposes several localizations, a frozen fixer scores each by how likely it makes a verified patch, and the localizer is trained on these scores with weighted supervised fine-tuning, without running tests and without reinforcement learning. We build the full pipeline on Gemma 4 E4B for the Kaggle Gemma-4 Developer Agent competition, from 1,629 test-verified Kimi K3 trajectories. The FGL target helps the fixer find the bug as often as the teacher's localization while using 27.5% less code. On the development set, one EM round moves the agent in the expected direction compared with a matched control: more tasks reach a patch (53 vs. 46), fewer tasks get stuck reading files that do not exist (43 vs. 54), and more tasks are solved (8 vs. 6). Training also changes how the agent works: trained agents explore about three times more than the base model and write far fewer wrong patches, and the single agent solves up to 11 of 78 tasks, against 8 for the base model. An analysis of 13 evaluation runs identifies what limits a 4B agent under an 8-minute budget (time, repetition loops and edit quality) and gives practical lessons for training small coding agents. All data, code and adapters are publicly available.
Read the full abstract

All authors Khalil Hennara, Muhammad Hreden, Zeina Aldallal, Sara Chrouf, Safwan AlModhayan

Statistical Methodology

Zero-inflated generalized linear latent variable models for microbiome compositional data

In this paper, we incorporate the Dirichlet–multinomial distribution into the generalized linear latent variable model (GLLVM) framework to accommodate high-dimensional, overdispersed compositional count data and capture complex covariation among taxa with a low-rank structure. Advances in high-throughput sequencing technologies have generated large amounts of microbiome count data, stimulating research into their associations with human diseases and traits. Such data are compositional and often highly sparse for various reasons, posing challenges for statistical analysis. Conventional zero-replacement approaches ignore the underlying causes of zeros, further complicating the analysis. We therefore propose a zero-inflated Dirichlet–multinomial latent variable model (ZIDM-LVM). Since the marginal likelihood has no closed-form expression, we develop a variational approximation (VA) approach for efficient parameter estimation. Simulation studies and applications to two microbiome datasets are presented to evaluate the performance of the proposed method.
Read the full abstract

All authors Wenqi Tang, Pekka Korhonen, Jenni Niku, Klaus Nordhausen, Sara Taskinen

Graphics

Same Physics, Different Gradients: A State-of-the-Art Report on Differentiable Physics Simulation

Differentiable simulation propagates derivatives through a physics simulator, turning animation, system identification, control, and computational design into gradient-based optimization. Engines differ in the physics they model and, less visibly, in which mathematical object they differentiate. We separate two questions the literature conflates. The referent is which object the returned derivative belongs to: the finite computation that executed, the converged equation it approximates, or the continuous dynamics. The realization describes the organization of the backward pass: automatically generated differentiation, an equation-level adjoint solve, or a custom adjoint. The referent defines the target against which exactness is assessed; the implementation determines how accurately that target is approximated. Cost depends on the realization together with how the states needed by the backward pass are retained or recovered. Crossing the axes places every method and engine we classify in a cell, separates claims the literature merges, and shows that the apparent trade-off between exactness and memory follows from how methods have been built rather than from a constraint. A second, orthogonal taxonomy covers contact, where the map is non-smooth. We then ask whether a correct gradient helps. Four mechanisms by which useful gradients are lost are cataloged and measured on one testbed, where the advantage of the exact derivative over the equation-level adjoint depends on where the fitting target comes from as much as on the solver budget, and we review the evidence that zeroth-order methods can outperform simulator gradients in the regimes where those mechanisms act. The model classes, systems techniques and applications are surveyed through this lens, and we close with a reporting checklist, a released diagnostic that computes it, and a research agenda.
Read the full abstract

All authors Lei Shu, Zizhou Huang, Yunuo Chen, Yifei Li, Tao Du, Minchen Li, Tuur Stuyck, Miles Macklin, Kui Wu, Bo Zhu, Yin Yang, Ming C. Lin, Alberta Longhini, Rika Antonova, Leonidas Guibas, Gordon Wetzstein, Jeannette Bohg, Chenfanfu Jiang

Cryptography and Security

GPU-Accelerated Non-Interactive Encrypted Neural Network Training with Batched Diagonal Packing

Training a neural network under homomorphic encryption requires transposes, batch reductions, optimizer updates, and repeated refreshes in addition to forward matrix products. We combine rectangular diagonal packing, feature-major SIMD mini-batch lanes, baby-step giant-step evaluation, and compact optimizer state in an MIT-licensed Rust CKKS implementation. Its CUDA path keeps training and bootstrap ciphertext arithmetic on the GPU, while retaining host key generation and client operations. In a matched 100-update experiment on the main non-interactive MNIST model, device-resident bootstrapping reduces training time from 14,268.295 to 293.075 seconds, a $48.68\times$ ratio against our CPU implementation, with exactly equal decoded parameters and predictions. Including setup and final evaluation gives a $40.67\times$ ratio. A separate GPU run completes ten passes over 4,800 images in 2,967.095 seconds and achieves 90.35% test accuracy versus 90.41% for its plaintext twin. These are single-initialization results; exact CPU/GPU parameter agreement is verified at 100 updates, not for the full 1,000-update run. Earlier completed CPU runs support a larger-batch key-switch advantage and baseline-batch accuracy within the registered tolerance, but the baseline-batch normalized key-switch target is missed. The released ReBoot baseline does not meet the common security-estimator threshold, and its full-setting CPU time is unmeasured. We therefore distinguish measured acceleration of our implementation from conditional layout comparisons and do not claim unconditional superiority over ReBoot.
Read the full abstract

Machine Learning

Beyond Flat Walks: Compositional Abstraction for Autoregressive Graph Generation

Autoregressive graph generators typically linearize graphs through flat node-edge walks, which can obscure higher-order structure in large and complex graphs. We introduce a hierarchical graph abstraction and coarse-to-fine generation pipeline that separates coarsening, tokenization, and decoding. This exposes three design axes: community coarsening, hierarchical tokenization, and order randomization. Across these axes, traversal-based hierarchy consistently improves distributional and structural fidelity, while its V.U.N.\ cost depends strongly on the coarsening and randomization policy. In particular, the fidelity–V.U.N.\ trade-off observed under within-community randomization is substantially reduced when traversal order is randomized across the full hierarchy. With a fixed backbone, training protocol, and matched training time across five molecular and five non-molecular benchmarks, hierarchical serialization yields consistent distributional and structural gains on all molecular datasets, while the non-molecular results further show that these gains depend on how well the supplied hierarchy aligns with structure in the target distribution. Together, these results characterize when and how hierarchical serialization benefits autoregressive graph generation.
Read the full abstract

All authors Kaiwen Bian, Andrew H. Yang, Yusu Wang, Ali Parviz, Gal Mishne

Quantum Gases

Scaling up spin complexity in ultracold assembled molecules: association and quantum chemistry of Cr-bearing diatomics

The availability of ever more atomic elements in the ultracold regime potentially enables the extension of molecule assembly well beyond bi-alkali systems to targeted chemical species for novel quantum science and technologies. The concurrent need for single-quantum-state control and molecular constants from ab initio calculations, especially for previously unexplored species, creates a trade-off between complexity and feasibility. Here, we address this challenge and exploit inter-isotope Feshbach spectroscopy as a tool to study three chromium-bearing diatomics ($^6$Li$^X$Cr, $^{50,52}$Cr$_2$, and $^{53}$Cr$_2$) at different levels of complexity. The extracted molecular properties, including the number of vibrational levels and the dispersion coefficients, allow us to assess quantum chemistry methods. Moreover, our understanding of atomic scattering allows us to assemble LiCr and Cr$_2$ molecules and clarify the observed association pathways with quantum state selectivity. Our results suggest unique prospects for the Li-Cr and Cr-Cr systems, and, more importantly, the experimental viability of a wider class of assembled high-spin molecules for many-body physics and precision measurements.
Read the full abstract

All authors Beatrice Restivo, Stefano Finelli, Andrea Simoni, Andreas Trenkwalder, Matteo Zaccanti, Alessio Ciamei

Earth and Planetary Astrophysics

Detecting space objects with Earth-observation radar satellites

Millimetre-to-centimetre orbital debris poses a hazard to spacecraft, but its abundance remains difficult to measure. Spaceborne radars offer a substantial sensitivity advantage at close range: received echo power decreases with the fourth power of distance, so an otherwise identical object ten times closer produces a 10,000 times stronger echo. We investigate this opportunity using openly available Sentinel-1 radar voltages recorded on six calendar days. Catalogue-assisted processing identifies a manually validated sample of 76 objects at ranges of 41.7–492.9 km, including satellites, rocket bodies and debris fragments. A blind search recovers two selected debris transits without using their orbital predictions during detection. A short debris encounter also demonstrates how precise range and Doppler measurements can refine a catalogue orbit. Ground scatter increases the effective background and is the principal sensitivity limitation in Earth-looking observations. An ideal receiver-noise sensitivity comparison predicts detectable conducting-sphere diameters of approximately 3.4 mm for Sentinel-1 and 0.6 mm for EarthCARE at 10 km range, assuming coherent integration limited by beam transit time and a 10 ms cap. These results establish the feasibility of space-object detection using existing Earth-observation radars and motivate statistical debris surveys and targeted orbit refinement following collision warnings. Observations directed away from Earth and receive windows adapted to nearby objects could improve sensitivity to small debris.
Read the full abstract

All authors Juha Vierinen, Laura-Kristin Scholtz, Ulrik Noer, Theresa Rexer, Daniel Kastinen, Pål Gunnar Ellingsen

High Energy Physics (Phenomenology)

FORM 5.0 and further developments

In these proceedings we review the features of the recently-released version 5.0 of the FORM computer algebra system, which is one of the core software packages for the large-scale symbolic manipulations required by calculations in theoretical high-energy physics. We also discuss developments since the release of version 5.0, focussing in particular on recent performance improvements.
Read the full abstract

Machine Learning (Statistics)

The balance between feature learning and collapse in dynamical systems-based generative models

Over the past decade, a growing body of work has established that implicit biases in learning dynamics fundamentally shape the solutions found by neural networks, governing their learned features and generalisation performance. However, despite these strides in deep learning theory, far less is known about representation learning in generative models based on dynamical systems, from normalising flows to diffusion. Importantly, empirical evidence suggests that continuous-time generative models can be prone to mode collapse, yet a learning theory explaining such biases remains elusive. To address this, we use tools from dynamical systems and operator theory to express the gradient flow of a general class of stochastic differential equations as a composition of linear operators, allowing us to relate collapse to low-rank biases in generative models. In variational inference and flow-matching tasks, we show that this bias can manifest in two different mechanisms: i) where collapsed models are stable fixed points of the gradient flow, or ii) in sequential feature learning, where smaller modes take progressively longer to be learned, producing an effective collapse for a fixed training budget. Together these findings point to a trade-off between feature learning and collapse, suggesting an operator view as a step towards a unified framework for understanding the learning dynamics of dynamical systems-based generative models.
Read the full abstract

All authors Julian Brandon, Bruno Loureiro, N Alex Cayco-Gajic, Arthur Pellegrino

Cryptography and Security

StealthyPoison: Manipulating LLM Agent Behavior through Procedural Poisoning of Third-Party Skills

LLM agents increasingly retrieve third-party skills from public registries and follow their procedures to complete user tasks. This reliance allows attackers to manipulate agent behavior by poisoning the skills. However, overtly malicious instructions can provide explicit signals for detection. We present StealthyPoison, a stealthy procedural poisoning attack that corrupts a skill's decision-critical rules to induce incorrect task outcomes while preserving its advertised functionality and normal execution. Unlike prior knowledge corruption attacks that craft poisoned content for known target question-answer pairs, StealthyPoison constructs reusable poisoned procedures from a publicly available third-party skill, without access to actual user requests or their ground-truth answers. To make the poisoned procedure generalize to unseen requests, it first synthesizes requests from the target skill and executes its clean version using a surrogate agent to obtain pseudo-references. These references and surrogate execution feedback guide the iterative refinement of payloads that alter decision-critical rules. To increase the likelihood that the poisoned skill is retrieved and selected over benign alternatives, the attack augments the poisoned skill's retrieval metadata with independently generated request descriptions. We evaluate StealthyPoison on 566 eligible target skills and 5,244 requests against more than 26,000 registry entries. It improves top-1 poisoned-skill placement by 13.6% relative to both HotFlip and GCG. Constructed using feedback from a single surrogate model, the procedural poisoning payloads transfer effectively and without modification to three held-out victim models and achieve end-to-end attack success in a Claude Code workflow powered by Opus 4.8. The defense evaluation further shows that most of these payloads evade detection by both SkillSpector and GPT-5.4-mini.
Read the full abstract

All authors Jingfen Qiao, Rui Li, Shuaiqun Pan, Olga Gadyatskaya, Andrew Yates, Evangelos Kanoulas

Optics

Silicon and silicon nitride photonic ring resonator arrays for photonic trimming metrology

Proposed silicon on silicon oxide 4x4 matrix multiplier for use in refractive index measurements, based on the Advanced Micro Foundry deep ultraviolet process. Agreement between simulated spectra and analytic design spectra demonstrate the intended performance of the components given a 1550 nm wavelength spectrum analysis. Intended purpose of the design includes optical damage analysis of the cladding, and photonic chip refractive index analysis following sub-threshold laser ablation. Femtosecond multi-shot tuning of refractive index is not a recommended application, due to round trip-losses causing quality factor and transfer function degradation.
Read the full abstract

Number Theory

Super-positivity via Dirichlet theta functions

Let $\chi$ be a quadratic Dirichlet character and let $\Lambda(\chi, s)$ be its associated completed $L$-function. The super-positivity property, introduced by Yun–Zhang (2017), asserts that $\Lambda^{(m)} \left( \frac{1}{2}, \chi \right) \geq 0$ for all $m \geq 0$. We show that a certain positivity property of its Dirichlet theta function $\theta(t, \chi)$, namely that $\theta(t, \chi)>0$ for all $t \geq 1$, implies the super-positivity property for $\chi$. We then investigate a refined form of the above property: we conjecture that the sequence $\{ \Lambda^{(2m)}\left(\frac{1}{2}, \chi \right )\}_{ m \geq 0}$ is always eventually increasing and has at most one turning point, and we prove these assertions again when $\theta(t, \chi)>0$ for $t \geq 1$.
Read the full abstract

Theoretical Economics

Individual incentives and uniform gains under symmetric constraints

We study the division of a fixed total under symmetric joint constraints when agents have continuous, strictly single-peaked preferences over their own shares. Each agent reports a complete preference, including comparisons across the two sides of her peak. For any finite population and any nonempty closed convex permutation-invariant feasible set, we characterize the uniform-gains rule by efficiency, symmetry, individual strategy-proofness, and own-coordinate peak-onlyness. The last condition requires that a change of preference leaving an agent's peak fixed also leave her own share fixed; it places no direct restriction on others' shares. The characterization separates uniqueness from continuity and yields a continuous characterization with individual incentives for compact feasible sets that are polyhedral or relatively full-dimensional and strictly convex. We then show that, for every population of two through six agents and every nonempty compact convex permutation-invariant feasible set, efficiency, symmetry, and individual strategy-proofness alone characterize uniform gains. The proof recovers allocations from the shares attainable by unilateral reports, feasible transfers, and endpoint reports. A common induction applies at each population, with additional starting configurations for five and six agents.
Read the full abstract

Rings and Algebras

On semigroups of order-decreasing partial isometries: presentations and ranks of certain ideals

Let $\mathcal{DP}_{n}^{-}$ denote the monoid of all order-decreasing partial isometries on a chain with $n$ elements, and let $\mathcal{ODP}_{n}^{-}$ denote the submonoid of $\mathcal{DP}_{n}^{-}$ consisting of all order-preserving partial isometries. In this paper, for each $1\leq r\leq n$, we determine minimal generating sets of the ideals $\mathcal{ODP}^{-}(n,r) =\{ \alpha \in \mathcal{ODP}^{-}_{n} : \lvert \im(\alpha) \rvert \leq r \}$ of $\mathcal{ODP}^{-}_{n}$ and $\mathcal{DP}^{-}(n,r) =\{ \alpha \in \mathcal{DP}^{-}_{n} : \lvert \im(\alpha) \rvert \leq r \}$ of $\mathcal{DP}^{-}_{n}$, thereby determining their ranks. Furthermore, we give presentations for the monoids $\mathcal{ODP}^{-}_{n}$ and $\mathcal{DP}^{-}_{n}$.
Read the full abstract

Computer Vision

EgoLabel: Zero-Shot Egocentric Auto-Labeling with Cross-Modal Consistency

Egocentric auto-labeling requires depth, camera pose, object, and human-pose estimates to agree across independently developed models. Differences in units, coordinate conventions, and scale can make individually plausible predictions geometrically inconsistent. We present EgoLabel, a zero-shot pipeline that combines off-the-shelf models without target-domain labeled training and provides explicit cross-modal registration and consistency diagnostics. The pipeline integrates MonST3R for geometry, LocateAnything with GMC-BoT-SORT and SAM~2 for objects, and V-HPOT for hands; FisheyeViT provides a separate whole-body compatibility demonstration. Our main hand evaluation uses H2O sensor depth. Across 20 sequences, per-hand depth anchoring reduces the mean of per-sequence median hand–depth residuals by $23\%$ relative to a single global scale, yielding a pooled median of $84.6\,\mathrm{mm}$ against a $20.5\,\mathrm{mm}$ ground-truth reference. Hand PA-MPJPE is $14.53\pm1.29\,\mathrm{mm}$, whereas absolute MPJPE is $119.1\pm31.3\,\mathrm{mm}$, revealing a substantial gap between aligned and metric accuracy. On three RGB-only sequences, camera ATE is $1.40\pm0.29\,\mathrm{cm}$ after $\mathrm{Sim}(3)$ alignment. We also identify a sliding-window implementation issue that allows sensor-initialized depth to change during optimization. Because hand anchoring and its residual share the same depth map, the residual measures internal consistency rather than independent accuracy. These results characterize both the utility and the current limits of the registration pipeline.
Read the full abstract

All authors Xiaozhi Zhang, Changjia Xu, Yaoyuan Yan, Haoxiang Jie, Bingjie Fu

High Energy Physics (Phenomenology)

SuperWIMPs in the Kination Era: Imprints on $N_{\rm eff}$ and CMB Spectral Distortions

We investigate how the late-time decays of unstable particles into dark matter and neutrinos can leave measurable imprints on both cosmological observables and collider experiments. We study this mechanism within the scotogenic model, where the lightest singlet fermion $N_1$ serves as the dark matter candidate and is produced via the late decay of the lightest inert doublet scalar via the superWIMP mechanism. By incorporating a kination-dominated era in the early Universe, the altered expansion rate allows to match the observed $N_1$ relic density in parameter regions that would be underabundant in standard cosmology. The energy exchange between the injected neutrinos and photons resulting from the late-time decay can produce CMB spectral $\mu$-distortions, detectable shifts in the effective number of relativistic species $N_{\rm eff}$, or both simultaneously. Concentrating on the lightest component of inert doublet scalars with masses between $200$ and $500~\text{GeV}$, we find that these cosmological signals are accompanied by complementary signatures accessible at the LHC. Although our analysis is derived within the scotogenic framework, our findings apply more broadly to scenarios in which the unstable particle undergoes late-time out-of-equilibrium decay to produce dark matter and non-thermal neutrinos, offering how future CMB measurements and collider searches can jointly probe dark matter production and the early Universe's thermal history.
Read the full abstract

Astrophysics of Galaxies

Intermediate-redshift ultra-luminous infrared galaxies as transitioning objects between main sequence and starburst galaxies

With their star formation efficiencies (SFE) and star formation rates (SFR) falling between normal star-forming galaxies and starbursts (SBs), ultra-luminous infrared galaxies (ULIRGs) at intermediate redshifts ($0.2<z<1$) have been postulated to be good probes of galaxies transitioning between these two regimes. Here, we present new Atacama Large Millimetre Array (ALMA) observations of a representative sample of 22 ULIRGs at redshifts $0.46<z<0.81$ evenly distributed across dust temperature ranges of $25<T_{\rm dust}<50~K$ and spanning infrared luminosities of $10^{12.0}<L_\text{IR}<10^{12.7}~L_\odot$. We detected CO(2-1) emission in 14 of these galaxies (64%). The SFE of these 14 targets are all similar, corresponding to depletion timescales on the order of hundreds of Myr. By employing two different versions of the main sequence relation of galaxies – one based on the specific SFR (sSFR) and one based on directly observable parameters (CO and IR luminosities) – we statistically confirm that the present sample of intermediate redshift ULIRGs indeed lies in the transitional region between MS and SB galaxies. In addition, we establish that the SFE of our sample of ULIRGs is a strong indicator of the position of a galaxy relative to the MS and the SB branches. Based on the observed morphologies and star-forming properties, we suggest that minor mergers provide a viable mechanism for temporarily enhancing the SFR and SFE of intermediate-redshift ULIRGs, although other evolutionary phases and pathways cannot be excluded.
Read the full abstract

All authors J.-H. Kim, D. Rigopoulou, M. Pereira-Santaella, L. Barrufet, A. Alonso-Herrero, I. García-Bernete, C. Pearson

Data Structures and Algorithms

Longest Unbordered Factors in Linear Time

A border of a string is a nonempty proper prefix that is also a suffix. A string is unbordered if it has no border. The length of a longest unbordered factor is closely related to the smallest period of the string. While the smallest period can be computed in linear time, the best previously known algorithm for computing a longest unbordered factor takes $O(n \log n)$ worst-case time for a string of length $n$. Whether this computational problem can be solved in worst-case linear time has remained open. In this paper, we resolve this open problem for integer alphabets of size $n^{O(1)}$ by presenting an algorithm that computes all occurrences of longest unbordered factors in $O(n)$ worst-case time. Our algorithm uses a combinatorial property of longest unbordered factors and scans the input string in both directions to capture all occurrences achieving the global maximum without computing the local maximum at every starting position.
Read the full abstract

Optimization and Control

Controllability with and without Dissipation - a Port-Hamiltonian Approach

A port-Hamiltonian system with energy dissipation can be modelled by connecting a larger, non-dissipative system with a so-called closure relation. We study whether approximative controllability of the dissipative system is characterised by the same property of the larger system. The answer is negative in general, but positive for sufficiently well-structured systems - this is our main result. As a simple illustration, we show how approximate controllability of a wave equation implies approximate controllability of a heat equation.
Read the full abstract

All authors Jochen Glück, Anthony Hastir, Birgit Jacob, Hannes Wagener

Computer Vision

DisasterVLG: Vision-Language Grounding and Evidence-Based Reasoning for Disaster Scene Understanding

General-purpose vision-language models struggle with crisis imagery due to its inherent noise, ambiguity, and high variability across disaster events - yet timely, accurate scene understanding is critical for humanitarian response. To address this, we introduce DisasterVLG, a large-scale multi-task dataset containing 1.24 million disaster images from 142 countries paired with 33 million structured annotations spanning dense captioning with localization, incident assessment, relational scene descriptions, and crisis-grounded visual question answering (VQA) across diverse hazard types and scene contexts worldwide. Unlike prior disaster benchmarks that focus narrowly on classification, DisasterVLG unifies recognition, grounded reasoning, and uncertainty-aware interpretation. A distinguishing feature is its evidence-grounded annotation design: each VQA instance includes supporting evidence text that links visual observations to answers, alongside positive and negative question pairs, multiple answer formats, and difficulty labels. Visual grounding layer links caption phrases to image regions and relationship statements to subject-object boxes, with explicit labels for visual supports and clues. To support rigorous evaluation, we construct a human-verified test set of 38K images (1.15M annotations) spanning broad geographic regions and disaster types. Experiments with domain-adapted vision-language models demonstrate strong performance across disaster captioning, incident recognition, and grounded VQA, while also exposing the limitations of general-purpose models in crisis-specific reasoning. DisasterVLG establishes a scalable, reproducible benchmark for vision-language models that must reason reliably under real-world uncertainty. We will publicly release the dataset, trained models, and code.
Read the full abstract

All authors Ali Diba, Aisha Al-Mohannadi, Yin Yang, Muhammad Imran, Ferda Ofli

Optimization and Control

On necessary conditions for feedback stabilizability when the control space is a fiber bundle

Within control theory, a question of great interest is to understand when a stabilizing feedback exists. Building upon Krasnosel'skiĭ and co-workers, necessary conditions for the existence of a continuous stabilizing feedback have been pioneered by Brockett and Coron, providing so-called adversarial and homological conditions, respectively. Kvalheim recently generalized their conditions from equilibria to compact attractors and—in line with earlier work by Coron—showed that when a control system can be identified with a vector bundle, the homology condition is stronger, that is, the homology condition implies the adversary condition. Kvalheim also showed that this comparison breaks when extending the case of vector bundles to fiber bundles. In this work, we propose and characterize an appropriate adversarial condition for control systems represented by fiber bundles. The homology condition, on the other hand, is already well-defined, showcasing its general applicability. With the generalized conditions at hand, we show that by appropriately relaxing the regularity of the feedback under consideration, that is, by allowing for mild discontinuities, the homology condition implies the adversarial condition for fiber bundles. Thereby, in this work, we extend these earlier comparative results.
Read the full abstract

Lattice Field Theory

Renormalized symmetry probes for emergent symmetries and non-chiral lattice fermions

The degeneracy of hadron correlators in symmetry-related channels signals symmetry restoration in high-temperature QCD. RG-invariant probes of this degeneracy need care when the regularization breaks chiral symmetry, or when the symmetry is not exact in the massless action, as for the emergent chiral-spin symmetry $SU(2)_{CS}$. We renormalize the thermal and screening meson correlators of every flavor sector by the thermal correlator of the same channel in the $\bar ud$ sector at a common reference point. The symmetry ratio $\kappa_{AB}$ then depends on the renormalization only through the ratio $R_{AB}$. For exact symmetries, $R_{AB}=1$ and $\kappa_{AB}$ is RG invariant and scheme independent in mass-independent schemes. For emergent symmetries, $R_{AB}\neq1$ carries the relative logarithmic cutoff dependence of the two channels, which cancels in $\kappa_{AB}$, and the residual dependence on the reference scale is physical. The framework applies to Wilson, twisted-mass and staggered fermions, where $R_{AB}$ absorbs the channel-dependent renormalization, and the tensor pair $(T_k,X_k)$ keeps $R_{T_kX_k}=1$ on isotropic Wilson-type lattices. Lattice artifacts are a small correction to the chiral-spin probes $\kappa_{CS}$ and $\kappa_{fade}$, but as large as the signal of the exact-symmetry probes. For free quarks, we derive $\kappa_{fade}=2/3$ for every mass and separation, and $\kappa_{CS}\le1/3$. We apply the renormalized probes to $N_f=2+1+1$ QCD with optimal domain-wall quarks at the physical point, with $a=0.064$ fm and $T=192$-$1539$ MeV. The hierarchical restoration of the chiral symmetries from the light to the charm sectors is confirmed, and is invariant under renormalization. The $\bar uc$ and $\bar sc$ sectors have the widest chiral-spin windows, also at a tighter threshold. Thus spin-1 mesons with one charm quark are the most attractive channels for the chiral-spin regime.
Read the full abstract