Robotics
HOTICE: Whole-Body Humanoid Object Transportation in Cluttered Environments
Toan Nguyen, Weiduo Yuan, Siheng Zhao et al. (+2 more)
Object transportation is a fundamental capability for
humanoid robots operating in real-world, human-centric environments, yet existing methods struggle when clutter constrains free space around both the
robot and its carried payload. We present HOTICE, a whole-body
humanoid learning framework for transporting objects through such cluttered environments. First, we introduce
Humanoid-Object Decoupled Potential Fields, which jointly encode collision-avoidance guidance for the
robot and the carried object, enabling coordinated, obstacle-aware motion for both. Second, to address the large action space inherent to whole-body loco-manipulation, we design a dual-
agent reinforcement learning architecture that decouples upper- and lower-body control while preserving whole-body coordination via shared state observations and rewards. To train a policy that generalizes across diverse cluttered scenes, we further employ a specialist-to-generalist
distillation strategy, in which privileged teacher policies are distilled into a single deployable student policy. We evaluate HOTICE in MuJoCo simulation and on a real Unitree G1
humanoid, demonstrating effective and robust object transportation across cluttered scenarios for objects of varying shapes. Our results show that HOTICE reliably coordinates whole-body motion and object-aware collision avoidance, generalizing effectively to previously unseen cluttered environments while achieving strong performance in sim2real deployment.
Read the full abstract ▾
All authors Toan Nguyen, Weiduo Yuan, Siheng Zhao, Yue Wang, Daniel Seita
Rings and Algebras
On multilinear polynomial identities for $2\times2$ matrices in characteristic $2$
Iritan Ferreira dos Santos
Let $K$ be an infinite field of characteristic $2$. It is well known that the Lie
algebra $\mathfrak{gl}_2(K)=M_2(K)^{(-)}$ of all $2\times 2$
matrices does not admit a finite basis of
polynomial identities. Similarly, the finite basis problem for the associative
algebra $M_2(K)$ remains open. In this note, we obtain two multilinear identities of degree $5$ for $M_2(K)$ and prove that, together with the standard
polynomial of degree $4$, they generate all multilinear identities of degree up to $7$. Moreover, we prove that none of the three generators can be omitted.
Read the full abstract ▾
Signal Processing
Beyond Legacy L1 C/A: Characterizing Advanced Multi-Frequency Spoofing Against Multi-Frequency GPS Receivers for Autonomous Navigation
Minhaj Uddin Ahmad, Muhammad Sami Irfan, Sagar Dasgupta et al. (+2 more)
Autonomous vehicles increasingly rely on high-end multi-frequency GNSS receivers that track modernized civil signals, such as L2C, L5, and L1C, alongside legacy GPS L1~C/A. Although multi-frequency operation improves robustness against natural error sources, its resilience against synchronous multi-band spoofing attacks remains insufficiently characterized. This paper presents a method for quantifying the inter-band
alignment accuracy required for a coherent spoofer to evade cross-frequency consistency monitoring that a resilient receiver may employ. As a foundational study, the presented method is implemented and evaluated using the GPS L1 C/A and L2C signal pair within a controlled software-defined environment. This environment encompasses deterministic signal synthesis, programmable inter-band code offsets realized through fractional-delay filtering, and processing by a dual-frequency GNSS-SDR receiver. An inter-frequency pseudorange-difference consistency detector is defined and calibrated using a constant-false-alarm-rate (CFAR) approach, yielding a threshold of approximately 19 m at a false-alarm probability of $10^{-3}$. Signal fidelity and receiver capture are verified through correlation-function analysis using a multi-tap correlator bank. The inter-band code offset is then swept from one code chip to 0.005 chips to characterize detector performance. Results show that a perfectly aligned coherent attack yields a detection probability of zero, while detection probability increases rapidly with inter-band misalignment, reaching approximately 50% at 0.05 chips (about 15 m) and approaching unity for offsets of 0.2 chips or greater. These results quantify the
alignment precision required for successful coherent dual-frequency spoofing and establish a practical framework for future evaluation of L5, L1C, and hardware-based multi-frequency spoofing scenarios.
Read the full abstract ▾
All authors Minhaj Uddin Ahmad, Muhammad Sami Irfan, Sagar Dasgupta, Mizanur Rahman, Thejesh N. Bandi
General Relativity and Quantum Cosmology
Operator-Level Quantum-Classical Correspondence in Relativistic Quantum Theory and Curved Spacetime
Pankaj Sheoran, Gopal Kashyap and Sanjay Siwach
In a recent work \cite{Shaikh:2026nsi}, it was shown that for a non-relativistic particle, the operator equations of motion can be cast in the same differential form as Newton's equation of motion. This establishes a correspondence between
quantum operator dynamics and classical mechanics at the level of the equations of motion. In this work, we investigate whether this correspondence extends to
quantum field theory and linearized
gravity. We explicitly show that the Heisenberg equations of motion for
quantum field operators reproduce the corresponding classical equations of motion for a broad class of field theories. The Klein-Gordon equation, Maxwell's equations, Yang-Mills equations, and linearized Einstein equations are shown to arise at the operator level without taking the limit $\hbar\rightarrow0$. We then extend the analysis to curved spacetime and show that the
quantum field operators satisfy the corresponding covariant field equations at the operator level. In the weak-field approximation, we show that the Heisenberg
evolution of the metric-perturbations reproduces the linearized Einstein equations at the operator level. For a
quantum particle in curved spacetime, the Heisenberg equations for position and momentum reproduce classical geodesic motion in the semiclassical limit, thereby extending the operator-level
quantum-classical correspondence from fields to particle dynamics.
Read the full abstract ▾
Quantum Physics
Superradiant Mpemba Relaxation in a Dicke Ladder
Matheus G. H. Santos, Hugo Sanchez, Italo M. de Araújo et al. (+3 more)
The Mpemba effect occurs when a state initially farther from stationarity overtakes a closer one during relaxation. We show that this anomalous ordering can coexist with superradiant emission in a collectively damped ensemble of two-level systems. In the symmetric Dicke
manifold, zero-temperature decay is a population cascade with rates $\Gamma n(N-n+1)$. We construct the sparse family $\rho_A=|\lceil2N/3\rceil\rangle\langle\lceil2N/3\rceil|$ and $\rho_B=(|0\rangle\langle0|+|N\rangle\langle N|)/2$. Although $A$ is initially more energetic and more distant from the stationary ground state, it crosses $B$ in both energy and trace distance because it starts in a more radiative region of the Dicke ladder. Both states develop emission peaks larger than their independent-emitter references. The crossing persists for all examined sizes from $N=6$ to $60$, while the peak intensities show an effective scaling close to $N^2$. The same pair has no crossing under independent decay, identifying collective radiative kinetics as the origin of the effect.
Read the full abstract ▾
All authors Matheus G. H. Santos, Hugo Sanchez, Italo M. de Araújo, Luis Felipe A. da Silva, Miled H. Y. Moussa, Norton G. de Almeida
High Energy Physics (Experiment)
Fast Timing at Colliders: From Scintillator Hodoscopes to MicroPattern Arrays
Sebastian N. White
We give a select overview of charged particle timing at
colliders. Roughly speaking, three eras capture the
evolution of timing arrays. While luminosity and multiplicity dictate the detector technologies, there is also a need to review the various approaches to signal capture and processing, and their role in changing conditions for calibration. For the current era–the High-Luminosity LHC upgrades–the technique of dual-threshold timing can significantly simplify the crucial correction for amplitude walk.This correction often spans a time range an order of magnitude larger than the target resolution. Furthermore, because operating conditions vary in the high radiation environment, calibration can no longer be thought of as a one-time procedure. The TOFHIR2C ASIC designed for a CMS upgrade can be configured to provide dual threshold timing. In this report, we use this capability to demonstrate the power of this technique suitable for our new era.}
Read the full abstract ▾
Superconductivity
Comment on "Andreev Reflection to Probe Momentum-Dependent Spin Polarization in Altermagnet CrSb"
Igor I. Mazin, Maxim Khodas and Boris Nadgorny
Zhang et al. [Phys. Rev. Lett. 137, 016701 (2026)] report momentum-dependent
spin polarization in the altermagnet CrSb from point-contact Andreev reflection along the (0001), (-1-120), and (10-10) directions. We show that the Andreev
spin polarization they claim to measure is forbidden by
symmetry: it vanishes identically in the diffusive regime for all directions in CrSb, and in the ballistic regime for two of the three orientations measured. It is allowed only along one direction, where the calculated band structure gives ~50%. The large, nearly isotropic values reported follow from an incorrect definition of ballistic Andreev
spin polarization that evaluates the
spin asymmetry at fixed three-dimensional momentum; the Andreev selection rule is local only in the momentum component parallel to the interface. Moreover, the ballistic regime is not demonstrated, and the fitting procedure is flawed.
Read the full abstract ▾
Computation and Language
TelecomGPT-R1: Unified Post-Training for Reasoning Across Heterogeneous Telecom Tasks
Bohao Wang, Chenwei Wu, Hang Zou et al. (+7 more)
Large
language models (LLMs) offer great potential to automate a broad range of telecom engineering tasks by
reasoning over standards, network configurations, mathematical models, source code, and operational logs. However, existing telecom LLMs struggle to reliably reason across these diverse tasks and data types. General-purpose LLMs often lack reliable grounding in telecom-specific knowledge, while telecom-specialized models are typically developed for narrower task families and exhibit limited multi-task performance. To fill this gap, we introduce TelecomGPT-R1, a family of open source unified telecom
reasoning models structured around four complementary axes: protocol, knowledge, modeling, and fault. We first develop an axis-aware data generation framework that refines coarse public telecom artifacts into verified question-answer pairs and high quality chain-of-thought (CoT)
reasoning trajectories, yielding a training corpus containing 104,880 examples. Building on this corpus, supervised fine-tuning (SFT) instills telecom knowledge and evidence-grounded
reasoning patterns to overcome the cold start barrier for reinforcement learning (RL). We then apply dynamic sampling policy optimization (DAPO) with task-routed rubric rewards to keep RL updates informative and stable across heterogeneous telecom
reasoning tasks. These rewards decompose axis-specific CoT traces into verifiable
reasoning units and combine grounded dense process credit with outcome correctness, allowing RL to learn generalizable problem solving behaviors from verifiable telecom evidence. We release the TelecomGPT-R1 models and a reproducible training recipe to support further community development. Evaluations on seven
benchmarks of the GSMA Open Telco Leaderboard show that the open-source TelecomGPT-R1-27B achieves an 89.64% mean score, outperforming leading proprietary models, including GPT-5, Claude, and Gemini.
Read the full abstract ▾
All authors Bohao Wang, Chenwei Wu, Hang Zou, Yu Tian, Lina Bariah, Li Wei, Chongwen Huang, Yongliang Shen, Zhaoyang Zhang, Merouane Debbah
Chemical Physics
MolExplain: An Interactive Tool for Explainable Molecular Property Prediction
Pirm Dhararag and Dylan Cashman
The application of machine learning to molecular property prediction has become increasingly prevalent in drug discovery, yet most models operate as black boxes, returning a prediction without revealing which structural features drive it. MolExplain addresses this gap by combining property prediction with sub-structure level visual explainability in an interactive web interface. The system featurizes molecules as Morgan fingerprints, classifies them using a trained XGBoost model, and applies SHAP attribution to produce a smooth heatmap overlay indicating which regions of the molecule contribute for or against the predicted property. Applied to cyclic peptide membrane permeability, the tool's attribution independently recovers the known role of backbone N-methylation in improving passive membrane
diffusion, consistent with established chemistry. While demonstrated on cyclic peptides, the framework is designed to generalize to other molecular properties, positioning MolExplain as a platform for interactive, explainability-driven molecular design.
Read the full abstract ▾
General Relativity and Quantum Cosmology
Thin-Shell Black Bounce
Leandro A. Lessa, Renan B. Magalhães and Gonzalo J. Olmo
A recently proposed class of black-bounce solutions sourced by anisotropic fluids, whose exterior reproduces the Reissner-Nordström geometry all the way down to a minimal-area bounce, was put forward by Lessa and Olmo. In this work, we show that this solution conceals a thin-shell structure precisely at the bounce. Motivated by this observation, we construct a new class of black-bounce solutions with static, spherical
symmetry featuring the emergence of a minimal-area bounce supported by a thin shell. This nontrivial topological structure is shown to arise from the degeneracy of the metric at the bounce, where the inverse metric becomes ill-defined, leading to
curvature tensors with distributional discontinuities precisely at that point. As a consequence, a delta-function thin shell of energy-condition-violating matter is necessarily present at the bounce. To generate such solutions, we draw on proposals from
quantum gravity that predict the existence of a fundamental minimum length scale, employing energy density profiles that encode these
quantum-gravitational effects in an effective manner, replacing the point-like description of matter at minimal-length scales with a smeared one. We consider two well-motivated profiles, of Lorentzian and Gaussian type, to generate a novel class of solutions exhibiting the features of either regular black holes or wormholes, whose center conceals a thin-shell bounce. In addition, we investigate the thermodynamics of these compact objects.
Read the full abstract ▾
Rings and Algebras
On the Peirce spectral rigidity of decorated incidence algebras
Daniel J. F. Fox and Vladimir G. Tkachev
The multiplication operator of an idempotent in a metrized commutative nonassociative
algebra is a self-adjoint linear endomorphism whose spectrum reflects the algebraic and
geometric structure encoded in the
algebra. Although for general
algebras the spectra of idempotents can behave arbitrarily, for the most interesting classes – among them
algebras of Clifford and Jordan type arising in geometry – the spectrum is highly constrained. We consider a family of
algebras determined by a partial Steiner triple system (PSTS) with blocks decorated by signs, each block determining an idempotent; even for the simplest PSTS the resulting
algebras can be quite complicated. For decorations of the Grassmannian PSTS $G_2(m)$, whose blocks are $3$-subsets of an $m$-element set and whose points parametrize an underlying incidence geometry, the spectra behave rigidly: the spectrum with multiplicity of a block idempotent is a function of a single combinatorial parameter, the number of negatively signed Pasch configurations – quadrilateral subconfigurations – containing the block. The resulting block spectra organize into a spectral profile, recovered by binomial inversion from a hierarchy of moment invariants. For a distinguished family of decorations, the associated
algebras are axial
algebras satisfying a fusion law independent of $m$, and for these
algebras there is given an alternative explicit
matrix model exhibiting the
rigidity directly; the even part of this model is a direct sum containing a polar
algebra summand, linking the construction to polar
algebras and symmetric Clifford systems.
Read the full abstract ▾
Cryptography and Security
SSP-Bench: A Hybrid Data Generation Framework for Safety, Security, and Privacy Evaluation
Fatih Deniz, Yazan Boshmaf and Issa Khalil
Evaluation of large
language models (LLMs) for safety, security, and
privacy (SSP) relies heavily on static
benchmarks, which suffer from score saturation, data contamination, and aggregation artifacts, and fail to capture sensitivity to linguistic variation. As a result, models that perform well on fixed test sets often fail under semantically equivalent rephrasings. We introduce SSP-Bench, a dynamic benchmarking framework that generates evaluation instances on demand while preserving domain consistency. The framework ensures label validity through externally grounded sources, enforces scope via service-specific validation, and calibrates difficulty using a multi-model steering panel.
Benchmark construction is formulated as a multi-objective optimization problem over difficulty, separability, novelty, and diversity. Across 24 models and four SSP services, SSP-Bench reveals systematic failures of static evaluation, including near-zero correlation in safety rankings due to construct mixing, strong safety–over-refusal coupling, and hidden within-family regressions. These results show that static
benchmarks can misrepresent model behavior, motivating dynamic, deployment-relevant evaluation.
Read the full abstract ▾
Robotics
Learning from Humans for Proactive Assistance in Human-Robot Collaborative Transport
Elvin Yang and Christoforos Mavrogiannis
We focus on human-
robot collaborative transport, a challenging task of broad relevance spanning logistics, manufacturing, and the home, in which a user and a
robot work together to relocate a large or heavy object. To act as an effective partner, the
robot should reduce the user's effort by contributing to efficient relocation of the object while remaining physically responsive to them. Prior work often addresses these capabilities separately, producing
robots that may move the object efficiently but resist user input, or accommodate the user but depend on continuous guidance. Our key insight is that obstacle-constrained collaborative transport requires integrating predictions of human collaborative behavior with compliant
robot control. To this end, we introduce PROACT, a framework for human-
robot collaborative transport that incorporates anticipation into compliant whole-body control through a learned model of human collaborative behavior. Trained on a large-scale, real-world
dataset of dyadic human transport demonstrations, our
transformer architecture distills collaborative behavior into predictions of future object motion. Across 108 real-world trials with a 9-DoF mobile manipulator, PROACT reduces mean interaction work by 59.2% and 20.4%, and mean completion time by 12.9% and 6.9%, relative to compliance-only and MPC baselines, respectively. Footage from our experiments can be found at https://youtu.be/qAGvQfVPjbk.
Read the full abstract ▾
Numerical Analysis
An unconditionally stable second-order Robin partitioned method for fluid-poroelastic structure interaction
Wenlong He, Thomas Wick, Xiaohe Yue et al. (+2 more)
We study an unsteady fluid-poroelastic structure interaction (FPSI) problem, in which the free fluid region is governed by the incompressible Stokes equations and the poroelastic region by the fully dynamic Biot system. The original physical coupling conditions are equivalently reformulated as Robin-type interface conditions and an auxiliary interface variable is introduced as common Robin data shared by the two subsystems. This reformulation decomposes the coupled FPSI problem into fluid and poroelastic subproblems posed on their respective subdomains. The problem is discretized in space by the finite element method (FEM), while the backward Euler scheme and the BDF2 scheme combined with the second-order interface extrapolation are employed for the temporal discretization. By exploiting the resulting discrete dynamic relations, we derive an update formula for the auxiliary interface variable and obtain fully discrete first- and second-order Robin partitioned schemes in which the two subproblems are solved independently at each time step. The unconditional stability of both schemes is rigorously established. Numerical experiments verify the temporal convergence orders of the proposed schemes and investigate the influence of the Robin parameter. A classical pressure-wave
benchmark verifies the accuracy of the proposed methods, while supplementary nonlinear moving-domain simulations illustrate their applicability to more complex FPSI configurations.
Read the full abstract ▾
All authors Wenlong He, Thomas Wick, Xiaohe Yue, Jiwei Zhang, Haibiao Zheng
Astrophysics of Galaxies
Calcium Triplet Absorption is Common around Little Red Dots
Jenny E. Greene, Hanpu Liu, Raphael E. Hviding et al. (+29 more)
We present new evidence for an optically thick atmosphere surrounding Little Red Dots (LRDs) in the form of Ca absorption at rest-frame 8500A, the calcium triplet (CaT). Building on the detection of CaT absorption in one local LRD analog (the "Egg", Lin et. al. 2025), we investigate the region around rest-frame 8500A for an archival sample of 17 LRDs ($2 < z < 5$) with JWST/NIRSpec grating data. We detect CaT absorption in three individual sources, and find that on average there is an absorption equivalent width (EW) EW$_{\rm CaT} \approx -5$ A. Among the three detections, two are approaching the deepest absorption seen in integrated light from
stars. Given that the continuum around CaT is probably dominated by the central engine, we argue that the absorbers are likely to be associated with the LRD. Such absorption has not historically been associated with any components of an AGN central engine, but a cool, optically thick photosphere as proposed for LRDs would naturally produce such absorption. The EW$_{\rm CaT}$ that we observe can be matched by hydrostatic atmosphere models at relatively low
metallicity ([M/H]$<-1$) combined with an effective temperature $T_{\rm eff} > 4500$ K. Alternatively, the distribution can be matched at a low photospheric gas density $\rho_{\rm ph}<10^{-11}{\rm~g~cm^{-3}}$ that requires a non-hydrostatic gas structure on dynamical grounds. In the future, metal absorption lines should be a powerful complementary probe of the gas conditions, and possibly the enclosed mass, of LRDs.
Read the full abstract ▾
All authors Jenny E. Greene, Hanpu Liu, Raphael E. Hviding, Rohan P. Naidu, David J. Setton, Bingjie Wang, Ivo Labbe, Vasily Kokorev, Xiaojing Lin, Hollis B. Akins, Anna de Graaff, Joel Leja, Jorryt Matthee, Alberto Torralba, Karl Glazebrook, Rachel Bezanson, Leindert A. Boogaard, Gabriel Brammer, Caitlin M. Casey, John Chisholm, Xiaohui Fan, Lukas Furtak, Seiji Fujimoto, Yan-Fei Jiang, Yilun Ma, Themiya Nanayakkara, Erica J. Nelson, Richard Pan, Eliot Quataert, Anthony J. Taylor, Christina Williams, Adi Zitrin
Mathematical Physics
Comparing the Lee-Yang gap with the mass gap in the sub-critical planar Ising model
Noam Borgnia, Jui-Hui Chung and Jacob Shapiro
We prove that the subcritical mass gap is comparable to the Lee-Yang gap over free-boundary square boxes raised to the power $8/15$. Random-current moment bounds, a correlation inequality and a finite block expansion give the upper bound. The lower bound follows from the Lee-Yang product, the susceptibility and the critical magnetization. We also obtain uniform factorial bounds on the field derivatives of the pressure.
Read the full abstract ▾
Digital Libraries
Open Science, Closed Models: How Funding Shapes AI in Science
Ana Trišović and Janakan Sivaloganathan
How funding shapes AI engagement in science is poorly understood despite its structural importance. We analyze 104,226 papers (2018-2025), linking funding acknowledgments to how each paper engages with foundation models: whether it extends a model (fine-tunes or builds on it), uses one without modification, or references models only peripherally. We report three findings. First, both private and public funding are associated with greater AI engagement. Public funding is associated with more off-the-shelf and open-weight use, while private funding is associated with greater model extension. Industry cloud credits are associated with less open-weight use, consistent with greater reliance on closed models. Second, industry involvement is associated with model extension, suggesting that corporate resources reach science through coauthorship, though this channel is concentrated in Computer Science and Global North collaborations. Third, the gap across research systems is not one of AI adoption but of control. Off-the-shelf use is rising across fields and country groups, and the least publicly funded group, the Global South, now has the highest share of API-only engagement (0.4% to 10.0% of engaged papers). China shows the opposite profile: the highest public-funding share and lowest API-only engagement, with extension remaining stable as it declines across the Global North. Standard capacity indicators show convergence in AI adoption even as dependence on proprietary infrastructure diverges.
Read the full abstract ▾
Rings and Algebras
Reflectors and the Globalization Problem for Partial Actions of Semiconstellations
Rafael Haag and Thaísa Tamusiunas
We introduce semiconstellations and restriction semiconstellations, together with their actions on sets, providing a common setting for partial actions of categories, inverse semigroupoids, restriction
semigroups, and semigroupoids. We then investigate the globalization problem, which asks whether every partial action can, up to isomorphism, be obtained by restricting a global action. A distinctive feature of our approach is the use of reflectors not only to construct globalizations, but also to study when partial actions are globalizable. We show that every partial action of a restriction semiconstellation admits a reflector in the full subcategory of $S$-
algebras consisting of what we call almost global actions, and investigate conditions under which this reflector provides a globalization. As an application, we obtain a positive answer to the globalization problem for partial actions of restriction semigroupoids on sets, and consequently for restriction
semigroups and restriction categories. In contrast, the answer is negative for restriction semiconstellations in general. Finally, given a semiconstellation $S$ and a suitable equivalence relation $R$ on $S$, we introduce the notion of an \emph{$R$-compatible action of} $S$ and construct a restriction semiconstellation $S^R$. We then show that partial and $R$-compatible actions of $S$ correspond, respectively, to partial and global actions of $S^R$. In this way, the globalization problem for semiconstellations can be reduced to the corresponding problem for restriction semiconstellations.
Read the full abstract ▾
Optimization and Control
Higher-Order Sensitivity Analysis of Wasserstein and Adapted Wasserstein Distributionally Robust Optimization Problems
Nathan Sauldubois
We investigate the higher-order expansions of the sensitivities of functionals of probability measures. Building on the work of \citeauthor{bartl2021sensitivity} \cite{bartl2021sensitivity} and \citeauthor{bartlsensitivityadapted} \cite{bartlsensitivityadapted}, who established the first-order expansion when the functional arises from a
stochastic optimisation problem, we extend these results using recent developments in differential calculus on Wasserstein spaces. Specifically, we generalise their approach to broader classes of functionals defined on the Wasserstein space. We provide an explicit formula for the second-order expansion of the sensitivity of a functional with respect to both the Wasserstein and the adapted Wasserstein metrics. Furthermore, we propose a construction method for the higher-order expansion of these sensitivities. Finally, following recent advances by \citeauthor{Touzisauldubois2024ordermartingalemodelrisk} \cite{Touzisauldubois2024ordermartingalemodelrisk} and \citeauthor{JiangObloj} \cite{JiangObloj} concerning sensitivity of Distributionally Robust Optimisation under martingale constraints, we also derive higher-order expansions for sensitivities in the presence of such constraints.
Read the full abstract ▾
Instrumentation and Methods for Astrophysics
Convolutional Neural Network for Extraction of $n = 1$ Photon Ring of Black Holes
Courtney Duong, Frank Myhre and Joseph R. Farah
Very-long-baseline interferometry (VLBI) may enable direct imaging of fine, event-horizon scale black hole structures such as the
photon ring. In particular, the $n=1$
photon subring encodes information about black hole
spin in its radial profile and azimuthal brightness modulation, making it a key proxy for spacetime properties. However, the $n=1$ subring overlaps with emission from the $n=0$ subring in observations, requiring isolation of the $n=1$ subring for measurements. We present a convolutional neural network (CNN) capable of extracting the $n=1$ subring from composite images containing $n=0$ and $n=1$ subrings. We simulate $111,000$ $m$-ring images with eht-imaging to train and validate the CNN. The CNN accurately recovers the radial and angular brightness profiles of the $n=1$ subring with a mean normalized cross correlation (NXcorr) of $\sim0.99$. The CNN provides a detection capability by predicting no false positives when input images lack a $n=1$ subring. We perform feature extraction with ringfit to show that the CNN-predicted $n=1$ subring accurately recovers the ground truth radial profiles and azimuthal brightness modulation, demonstrating the CNN's capability to extract
spin-sensitive features. We further test the CNN on black hole images generated by KerrBAM and general relativistic magnetohydrodynamic (GRMHD) simulations, finding that it recovers the overall $n=1$ subring radial profiles while exhibiting discrepancies in the recovered intensity profiles. These results demonstrate the potential of deep learning methods to isolate the $n=1$ subring from overlapping emission, providing a framework for analyzing future high resolution black hole images from proposed VLBI missions such as the Black Hole Explorer (BHEX).
Read the full abstract ▾
Solar and Stellar Astrophysics
Asteroseismology and interferometry of the F7V spectroscopic binary $\chi$ Draconis A in the TESS CVZ
Jonatan Rudrasingam, Mikkel N. Lund, Frank Grundahl et al. (+26 more)
We present a detailed analysis of the asteroseismic main-sequence
benchmark star $\chi$ Dra A in the TESS northern CVZ. We aim to derive
stellar mass and radius from asteroseismic modelling of individual mode frequencies and test the asteroseismic mass and radius against an independent dynamical mass and interferometric radius measurement. We determined the dynamical mass of $\chi$ Dra using 618 radial velocity measurements obtained with the SONG
telescope at Tenerife, and 53 relative astrometric measurements. With the PAVO beam combiner at CHARA, we obtained the interferometric radius of $\chi$ Dra A. We determined asteroseismic parameters from 16 sectors of 20-sec cadence TESS photometry. We determined $T_{\rm eff}=6277\pm30$ K, $\rm [Fe/H] =-0.51\pm0.03$ dex, and $[\alpha/\rm Fe]=0.08\pm0.03$ dex from the spectroscopic analysis. For the dynamical fit we obtained a mass of $M_{\rm A}=1.028 \pm 0.004$ $\rm M_{\odot}$ for $\chi$ Dra A, and $M_{\rm B}=0.735 \pm 0.003$ $\rm M_{\odot}$ for $\chi$ Dra B. Combining the derived interferometric angular diameter with dynamical parallax yields an interferometric radius of $R_{\rm A} = 1.159^{+0.029}_{-0.028}$ $\rm R_{\odot}$. In the TESS power spectrum, we identified 38 individual oscillating modes. Using these modes, we modelled the
star with nine independent pipelines to test the resulting model mass and radius against our independently calculated mass and radius. All models yielded masses slightly lower than the dynamical mass. Using asteroseismic scaling relations, we found that scaling masses from corrected scaling relations best reproduce the dynamical mass. The combination of spectroscopy, interferometry, and asteroseismology has yielded precise results for the main component of the $\chi$ Dra system, making it one of the best-characterised main-sequence
solar-like oscillators.
Read the full abstract ▾
All authors Jonatan Rudrasingam, Mikkel N. Lund, Frank Grundahl, Arnab Chowhan, Timothy R. Bedding, Poul E. Nissen, Aldo Serenelli, Joel M. J. Ong, Jeppe S. Thomsen, Yaguang Li, Eva Panetier, Sarbani Basu, Owen J. Scutt, Travis S. Metcalfe, Savita Mathur, Mark Lykke Winther, George T. Hookway, Benjamin J. S. Pope, Amalie Stokholm, Desmond H. Grossmann, Jens R. Larsen, Jørgen Christensen-Dalsgaard, Daniel Huber, Mads S. Fredslund, Timothy R. White, Paul G. Beck, Pere L. Pallé, Rafael A. García, Hans Kjeldsen
High Energy Astrophysical Phenomena
Multi-scale variability in the optical afterglow of GRB 251013C from high-cadence LAST observations
R. Konno, E. O. Ofek, S. Garrappa et al. (+8 more)
Context. Optical afterglows of gamma-ray bursts (GRBs) are often described by smooth power-law decays, but deviations such as
flares and rebrightenings provide important constraints on the underlying emission processes. Aims. We investigate the temporal variability of the optical afterglow of GRB 251013C using high-cadence observations, and assess the origin of the observed variability. Methods. We analyze continuous optical observations obtained with the Large Array Survey
Telescope (LAST), complemented by publicly available X-ray data from Swift-XRT. The optical light curve is modeled using a power-law decay with superposed variability components, and contemporaneous optical and X-ray measurements are used to constrain the broadband spectrum. Results. The optical light curve shows pronounced variability on multiple timescales, including broad rebrightening episodes with $\Delta t \sim t$ and superposed faster fluctuations with $\Delta t/t < 1$. The main rebrightening exhibits a structured rise with a clear steepening prior to the peak. The fast variability comprises asymmetric fast-rise, slow-decay features whose durations increase with peak time. The optical flux lies on the extrapolation of the X-ray spectrum, and the spectrum hardens by the same amount whether measured within the X-ray band or from optical to X-ray. Conclusions. We favor a refreshed external-shock interpretation for the main rebrightening. The broadband spectrum and post-peak decay are consistent with slow-cooling synchrotron emission in a wind-like medium. The spectral hardening requires a newly dominant, harder electron population, while the faster optical variability indicates structure within the refreshed ejecta or shocked region.
Read the full abstract ▾
All authors R. Konno, E. O. Ofek, S. Garrappa, O. Teboul, E. Waxman, S. Ben-Ami, D. Kovaleva, A. Krassilchtchikov, D. Polishook, E. Segre, E. A. Zimmerman
Number Theory
Distinct exponents in the prime factorization
Mikhail R. Gabdullin and Vitalii V. Iudelevich
Following Erdős (1982) and Sanna (2019), we study the arithmetic function $h(n)$, which is defined to be the number of distinct exponents in the
prime factorization of a positive integer $n$. Among other things, we show that $$ \sum_{n\leq x}h(\phi(n)) \asymp x\left(\frac{\log\log x}{\log\log\log x}\right)^{1/2}, $$ where $\phi$ is the Euler totient function. The key ingredient is the Poisson random model for $\omega(n,T)$, the number of the
prime divisors of $n$ in a given subset of
primes $T$, which was introduced in a recent work of Ford.
Read the full abstract ▾
Machine Learning
Concept Drift from a Causal Perspective
Eduardo V. L. Barboza, Jean Paul Barddal, Robert Sabourin et al. (+1 more)
Concept drift is a common phenomenon in real-world data streams, in which changes in the data-generating distribution can degrade predictive model performance. Most existing definitions characterize drift as changes in the joint distribution $P(\mathbf{x}, y)$, without distinguishing which component of the data-generating process has changed. In this work, we introduce a causal perspective on concept drift based on Structural Causal Models (SCMs). We propose a taxonomy that categorizes drift events by their causal origin, including changes in exogenous variables, endogenous mechanisms, confounders, and target-generating processes. Building on this framework, we develop an SCM-based data stream generator that simulates controlled mechanism-level drift events. Our experiments empirically characterize the distributional effects of each drift type and show that drifts with different causal origins induce distinct patterns of distribution shift and predictive behavior. Furthermore, by integrating causal discovery methods, we use our framework to construct data streams grounded in real-world dependency structures, enabling more realistic and informative evaluation scenarios. We also demonstrate that leveraging the generated data can improve downstream performance. These results highlight the importance of accounting for causal structure when studying and evaluating adaptive learning methods, and establish a foundation for causally-aware evaluation in non-stationary environments.
Read the full abstract ▾
All authors Eduardo V. L. Barboza, Jean Paul Barddal, Robert Sabourin, Rafael M. O. Cruz
Earth and Planetary Astrophysics
A Very Fast Symplectic Integrator for Planetary Systems: WHFast512 in x86 Assembly
Rishit Dagli and Hanno Rein
Long-term integrations of planetary systems are used in a variety of astrophysical applications, for example to determine a system's stability. In this paper, we describe significant speed and accuracy improvements to the WHFast512 integrator. By completely rewriting WHFast512 in x86 assembly, we gain fine-grained control over CPU instructions and are able to keep the entire simulation state in CPU registers, only writing to
memory when an output is required. We also reduce the number of branches in the code, improve the Kepler solver's speed and accuracy with a convergence check, switch the
Hamiltonian splitting to Jacobi coordinates, implement
symplectic correctors, and add support for ejection and collision detection. As a
benchmark we run simulations of the
Solar System with all eight
planets and general relativistic corrections for 5 Gyr. With our improvements, we can integrate this system in 9 hours, more than 8x faster than the standard version of WHFast. This makes WHFast512 by far the fastest N-body integrator for this kind of simulation. We run extensive tests to verify the integrator's accuracy, with a particular focus on removing any long-term bias, and show that the accuracy is now five orders of magnitude better for typical setups compared to the earlier version of WHFast512. The new WHFast512 integrator is freely available in the REBOUND integrator package.
Read the full abstract ▾
Robotics
GINIO: A Geometric SO(3)-Equivariant Interface for Neural Inertial Odometry
Chankyo Kim, Minghan Zhu, Tzu-Yuan Lin et al. (+2 more)
Neural inertial odometry increasingly uses networks as learned measurements inside filtering pipelines. Such measurements should transform consistently under arbitrary IMU mounting conventions: their mean must transform as a vector, and their covariance must transform congruently as a second-order tensor. We present GINIO, a
geometric SO(3)-equivariant interface for neural inertial odometry under arbitrary rotations of the IMU measurement frame. Given calibrated IMU windows, our framework predicts a motion measurement and uncertainty obeying these tensorial laws. To support efficient sensor-frame learning, we introduce Last-Frame
Alignment (LFA), a deterministic preprocessing step that is provably equivalent to world-frame training for SO(3)-equivariant predictors. The connected
estimator tracks sensor-local states such as IMU bias, separating nuisance estimation from the
geometric law enforced by the learned measurement. We instantiate the same interface in filter-connected NIO, AirIO-style recurrent aerial prediction, EqNIO-style full-SO(3) canonicalization, and ResNet-style temporal backbones. On TLIO, GINIO achieves 2.018 m ID/SO(3) ATE while EqNIO degrades to 76.389 m, using 11.6x fewer FLOPs. On NanoBench, our AirIO-style instantiation improves ATE from 5.579 m to 1.430 m without external attitude input, and our ResNet-style instantiation reaches 0.581 m ATE versus 0.645 m for ResNet1D. On Fetch, GINIO empirically reduces unseen physical-remount ATE from 8.15 m to 0.50 m without retraining, demonstrating robustness beyond the exact coordinate-frame guarantee. For uncertainty, spectral covariance reduces covariance-equivariance error by over three orders of magnitude compared with a diagonal head.
Read the full abstract ▾
All authors Chankyo Kim, Minghan Zhu, Tzu-Yuan Lin, Avantika Rattan, Maani Ghaffari
Artificial Intelligence
Clarification Is Not Correction: LLMs Fail to Let Go
Jianzhe Lin, Xiaolin Li, Fei Wang et al. (+3 more)
Dialogue failures in
language models are usually framed as
memory failures: context too long, summaries lossy, a constraint forgotten. We argue this misses a deeper problem: in many conversations the model does not forget, it commits too early. An ambiguous early turn collapses into a single hidden interpretation, and later clarification is filtered through that commitment. We call this early posterior collapse: unresolved user intent collapsing into a committed task state before ambiguity is resolved. We study it with controlled dialogue tasks in writing, planning, and coding using Gemini-2.5-Pro and Gemini-2.5-Flash. Across thousands of trials, the same information in different orders yields different outcomes, even when the final dialogue contains equivalent task-relevant information. This order effect suggests later clarification is treated as extra context rather than a corrective signal: it refines a stale task state without invalidating it. Coding tasks are especially vulnerable, suggesting early assumptions get embedded in structured artifacts such as interfaces and control flow. Standard prompting and
memory strategies do not reliably help: summaries can collapse ambiguity, and chain-of-thought can reduce explicit wrong commitment in
reasoning traces without improving final task success. These findings motivate uncertainty-preserving state management. If assistants cannot let go of early interpretations, robustness cannot rely on post hoc correction alone; it must keep ambiguous early turns from hardening into one task state. Assistants should hold tentative hypotheses while ambiguity remains, ask before executing when high-impact ambiguity persists, and rebuild from a revised state when later evidence invalidates an earlier reading. Rather than one prompting fix, we aim to redirect research for interactive LLMs from retaining more context toward preserving uncertainty.
Read the full abstract ▾
All authors Jianzhe Lin, Xiaolin Li, Fei Wang, Robert Douglas, Rajeshkumar Golani, Jubin Chheda
Quantum Physics
Thermalization and dephasing in an isolated system of coupled qubits
Jukka P. Pekola and Bayan Karimi
A system evolving unitarily does not dephase or thermalize by definition. Yet, if one considers a part of a
unitary system, situation is different. Here we analyze an isolated set of coupled
qubits taking one
qubit at a time as a subsystem, and writing self-consistently a weak-coupling master equation for it. Weakly anharmonic oscillators and ideal
qubits (two-level systems) demonstrate dephasing, i.e. the diagonal elements of the density
matrix reach a steady-state in the long-time limit, and the off-diagonal elements of ideal
qubits decay exponentially. For the linear coupling between the
qubits, only the degenerate ones interact, and the system does not reach thermal distribution. However, non-linearity, in our analysis in form of three-wave mixing, ensures a thermal distribution in the long time limit, where the temperature is determined uniquely by the energy in the initial non-thermal state. Finally, we present a potential experimental scheme based on a superconducting
quantum circuit.
Read the full abstract ▾
Hardware Architecture
GRADE-RTL: Evaluating LLM-Generated RTL Beyond Compilation
Hepziba Susan, Shivaranjani G. R., Malik Imran et al. (+3 more)
Large
language models (LLMs) can generate register-transfer-level (RTL) code from natural-
language specifications, but compilation alone does not establish structural completeness, functional correctness, or implementation efficiency. This paper presents a framework for evaluating LLM-generated RTL beyond compilation, which we named GRADE-RTL. Using fixed
prompts and validation settings, GRADE-RTL applies five checks, such as Port Signature, Compilation, Elaboration, Module Completeness, and Functional Equivalence against trusted reference RTL. With the help of compact metrics, we identify where candidates fail and distinguish end-to-end success from eligibility for downstream synthesis. We evaluate nine general-purpose and RTL-specialized LLMs on ten edge-relevant intellectual property designs under a budget of three generation attempts with failure-directed feedback. End-to-end success ranges from 0% to 70% across the evaluated models, with failures extending beyond compilation to hierarchy resolution, incomplete logic, and behavioral mismatch. FPGA implementation and 65 nm ASIC synthesis results further show that functionally equivalent RTL can differ substantially in resource use, timing, area, and power. A PID-controller place-and-route case study illustrates the physical-design consequences of different RTL implementations. GRADE-RTL provides a practical basis for comparing LLM-generated hardware descriptions by separating structural validity, behavioral correctness, and implementation quality.
Read the full abstract ▾
All authors Hepziba Susan, Shivaranjani G. R., Malik Imran, Muhammad Rashid, Sumathi Gokulanathan, Zain Ul Abideen
Machine Learning
MT-ProtBERT: Multi-task Learning ProtBERT for Intrinsically Disordered Proteins Classification with Scarce Data
Jian Sun, Kingshuk Ghosh, Lilianna Houston et al. (+1 more)
Intrinsically disordered
proteins (IDPs) differ from folded
proteins in that they are dynamic, lack a stable three-dimensional conformation, and have low sequence similarity between similar
proteins. The conformational heterogeneity of IDPs - while beneficial for their diverse functions - limits the use of traditional experimental tools to determine their conformation. The experimental difficulty, along with low sequence similarity, results in data scarcity, and makes it difficult to classify/detect IDPs that are similar or dissimilar, a task relevant to understand biology and
evolution. We address this challenge using Multi-task ProtBERT (MT-ProtBERT), a multi-task extension of ProtBERT tailored for low-data regimes. MT-ProtBERT integrates Dynamic Window Masking, a Multi-Scale 1D Convolutional classifier (MS-Conv1D), and auxiliary objectives that jointly optimize masked
language modeling and biochemistry-informed tasks. We evaluate this framework on two tasks under limited data: (i) phosphorylation site prediction (S/T/Y) in short sequences and small
datasets, and (ii)
protein compaction prediction on two small
datasets (684 and 530 sequences), including sequences comparable in length to typical disordered regions. MT-ProtBERT consistently outperforms PARROT, an RNN-based IDP-specific model, across all tasks. These results demonstrate that combining self-supervised and biochemistry-informed tasks, and multi-scale learning enables robust modeling of unstructured
proteins under data scarcity.
Read the full abstract ▾
All authors Jian Sun, Kingshuk Ghosh, Lilianna Houston, Mohammad H. Mahoor