Robotics
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Friday 6 November
Humanoid whole-body control requires a robot to track diverse reference motions while maintaining balance under gravity, torque limits, contact switching, and external disturbances. Existing reinforcement-learning-based trackers typically learn control outputs that jointly address motion tracking and balance. However, similar reference poses can correspond to substantially different dynamic demands, requiring a unified policy to learn motion-dependent control requirements alongside balance corrections. This can complicate learning and lead to high-frequency fluctuations in control outputs. We propose DKWBC, a dynamics-informed whole-body control framework that combines explicit model-based tracking with a learned balancing residual. Inspired by the model-based and environmental-wrench contributions to joint torques in floating-base dynamics, DKWBC uses inverse-dynamics feedforward and joint PD feedback to supply tracking torques, while a residual policy learns additional joint-torque corrections for balanced execution. By explicitly accounting for motion-dependent torque demands, this structure simplifies the policy's learning task. Experiments in simulation and on real hardware demonstrate that a single learned balancing policy generalizes across motion datasets while maintaining accurate motion tracking and balance across ground support, attached payloads, and physical interaction. Project page: https://robbyant-research.github.io/DKWBC/.
Quantum Physics
Quantum subsets discrimination and key sifting with success probability one
Non-distinguishability of non-orthogonal quantum states is the essential feature of quantum mechanics. In particular, one cannot deterministically orthogonalize the set of quantum states, while probabilistic orthogonalization is possible, including unambiguous state discrimination and discrimination between the subsets. Here, we propose a method for error-free subsets discrimination with significantly increased success probability, which may reach one even for non-orthogonal and linearly dependent input states. This method may be applied in quantum cryptography: An alternative basis reconciliation procedure is described, which may provide unit sifting probability even for a large number of bases and for non-orthogonal states within each basis.
Read the full abstract
Algebraic Geometry
Simply connected quotients of fake quadrics
Fake quadrics are minimal complex surfaces of general type with $b_1=0$ and $e=4$. The paper gives a short introduction to their arithmetic construction, aimed at low-dimensional topologists, and studies their holomorphic and antiholomorphic quotients and their fundamental groups. The running example in the paper comes from a quaternion algebra over a totally real cubic field. The lattice is generated by three explicit pairs of matrices. A holomorphic involution gives (after resolving the singularities) a simply connected minimal complex surface of general type with $p_g=0$ and $K^2=4$, a smooth $4$-manifold homeomorphic but non-diffeomorphic to $CP^2$ blown up $5$ times. An antiholomorphic involution gives a smooth four-manifold homeomorphic to $S^4$, with a knotted torus as branch locus.
Read the full abstract
Rings and Algebras
An extension of the notion of algebraic entropy
The definition of algebraic entropy presents a difficulty in the general setting of filtered algebras, which makes it necessary to distinguish between filtered algebras and finitely filtered algebras. While algebraic entropy is well defined in the finitely filtered case, extending this notion to arbitrary filtered algebras is not immediate. To address this issue, we realize a filtered algebra as the limit of a strict direct system of finitely filtered algebras and define its entropy as the supremum of the entropies of the algebras in the system. This extension is consistent with the original definition, when the algebra is finitely filtered, the extended entropy agrees with its algebraic entropy. This new setting allows us to extend the definition of the algebraic entropy to path and Leavitt path algebras of infinite graphs.
Read the full abstract
Data Structures and Algorithms
A Fast and Accurate Algorithm for the $p \to q$ Norm of Sparse Non-Negative Matrices
We present an algorithm that computes the $p \to q$ norm of a non-negative matrix with $s$ nonzero entries to accuracy $\varepsilon$ in time $\widetilde{O}(s^{3/2} \log (1/\varepsilon))$, for any $p \ge q \ge 1$. The fastest algorithm previously designed for this problem has running time $\widetilde{O}(s / \varepsilon)$ (Objois and Vladu, 2026), which is slower when high accuracy solutions are desired. Our algorithm employs a surprising application of Laplacian solvers inside an Interior Point Method (IPM). We formulate the problem as a linear optimization problem over a product of power cones, and show that the Newton updates of the IPM can be computed by solving a Laplacian system in a Schur complement after subtracting a rank-2 matrix. We also discuss an implementation of our algorithm, showing that it is practical and it outperforms optimized solvers such as MOSEK by factors of up to $20\times$ on a variety of tasks.
Read the full abstract
Data Structures and Algorithms
Coresets for Capacitated Clustering via Dual Concentration
We study coresets for clustering problems, focusing on the classical $k$-median and $k$-means objectives. A coreset is a small weighted summary of the input data that approximately preserves the clustering cost for every choice of centers. Such summaries are a central tool for scaling clustering algorithms to large datasets. In this work, we consider the capacitated setting, where the clustering is subject not only to choosing centers but also to respecting prescribed cluster-size constraints. This model is a natural generalization of standard clustering and has been widely studied, but its coreset complexity is less well understood. For finite metric spaces, we obtain optimal-size coresets; surprisingly, these match the best-known bounds for unconstrained clustering. We also give improved coreset bounds in Euclidean space, notably the first coresets with near-linear dependence in $k$ while being independent of the dimension and input size. The main new technical ingredients are concentration bounds for the dual solutions of min-cost flow problems, which allow for both significantly simpler and tighter analyses of the underlying sampling distribution used in the coreset construction.
Read the full abstract
Computer Vision
Dual-Region Slice Aggregation for CT-Based Renal Tumor Classification
Clear cell renal cell carcinoma (ccRCC) is the most common kidney cancer in adults. We present a region-aware architecture for CT-based ccRCC classification. Tumor and 0–10 mm peritumoral context are pooled separately from shared frozen DINOv3 feature maps. A shared position-aware slice aggregator precedes fixed training-only principal component analysis (PCA). The resulting representations are fused with radiomics using two scalar feature gates. Evaluation on KiTS23 and 111 external cases yields internal test and external AUCs of 0.8078\pm0.0154 and 0.9094\pm0.0015. Compared with matched single-tumor readout, mean AUC increases by 1.50 and 2.34 percentage points, respectively. Early-PCA and reconstruction controls yield lower internal mean AUCs, while external AUC is similar for early and late PCA. These findings highlight the joint role of regional readout and slice aggregation in adapting frozen visual features to renal CT.
Read the full abstract
Machine Learning
Transformer-FSQ: Discrete Latents for Extreme KV-Cache Compression
Private, personalised language models must run on local consumer devices, where large, frontier models cannot fit, and retaining long contexts is non-trivial. Sequential memory—the key-value (KV) cache—grows linearly with context length, quickly dwarfing the model itself: for Qwen3-1.7B-Base, 200,000 retained entries require 21 GiB, 6.6 times more than the model weights. Yet the KV-cache compression methods we measured fail beyond moderate compression ratios. We introduce Transformer-FSQ, addressing extreme KV-cache compression by replacing standard key and value projections with a finite scalar quantisation bottleneck. Our cache stores a packed vector of discrete integer codes per token, with no per-token quantisation metadata. Converting a pretrained transformer trains only these projections on 0.25 tokens per original parameter. We evaluate Transformer-FSQ against five baseline families on models from 0.6B to 8B parameters. On long-context retrieval (RULER), from 12$\times$ to 97.5$\times$ Transformer-FSQ is the most accurate method we evaluate, across scales. A fused Triton kernel decodes directly from the integer codes; with baselines given their authors' kernels where available, it decodes at $1.01\times$ (95% interval $0.98$–$1.05$) the throughput of native SDPA attention on the uncompressed Qwen3-8B-Base at $32\times$ and $1.54\times$ at $97.5\times$. Extreme KV-cache compression is therefore a practical way to extend how much context a fixed device can hold.
Read the full abstract
Computation and Language
RuACT: Recovering Inactive Rubrics as Adaptive Critics and Teachers
Rubric-based reinforcement learning (RL) extends RL with verifiable rewards to open-ended tasks by scoring each response against query-specific rubrics. Under group-relative optimization, however, a rubric that every sampled response fails (unreachable) or passes (saturated) gives all responses the same outcome and no learning signal; in Qwen3-4B medical training, 60.88% of query-rubric observations are inactive. We find that the two states lack different supervision: stating an unreachable rubric in the prompt raises its satisfaction from 1.13% to 91.25%, whereas responses that all pass a saturated rubric still differ in quality. We therefore propose RuACT (Rubrics as Adaptive Critics and Teachers), which assigns each inactive rubric the role that supplies what it lacks. Rubric as Teacher uses unreachable rubrics as privileged information for reward-gated on-policy self-distillation, and Rubric as Critic decomposes saturated rubrics into child rubrics that separate current responses while sharing the parent's weight. RuACT outperforms GRPO, rubric reweighting, and rubric-guided exploration on all four medical and scientific benchmarks with Qwen3-4B and achieves the best average with Qwen3-1.7B. On initially unreachable rubrics, unguided recovery rises from 35.02% to 42.86%, while child refinement improves reward-quality alignment on independent rollouts from 57.74% to 60.77%. Code is available at https://github.com/ZJU-REAL/RuACT.
Read the full abstract
Machine Learning
Coordinatewise Epsilon for Adam under Heavy-Tailed Noise
Adam estimates second moments coordinatewise, but typically adds the same scalar $\varepsilon$ to every denominator. We study how this shared $\varepsilon$ affects convergence when noise and curvature vary across coordinates. Under coordinatewise finite-$p$ noise moments, $p\in(1,2]$, choosing $\varepsilon_i$ from the noise and curvature bounds for each coordinate attains the heavy-tailed $\ell_1$-stationarity benchmark $O(T^{-\frac{p-1}{3p-2}})$, matching the known optimal time exponent for smooth nonconvex stochastic optimization. For $1<p<2$, Adam with a shared $\varepsilon_t$ admits a lower bound of order $\Omega(T^{-\frac{p-1}{2p}})$. This lower bound holds for every predictable scalar choice of $\eta_t$, $\beta_{1,t}$, $\beta_{2,t}$, and $\varepsilon_t$, precluding a dimension-uniform benchmark rate for this class.
Read the full abstract
Quantum Physics
Modular Hamiltonian and Typicality of Degenerate States
Free fermions exhibit Eigenstate Typicality (ET) meaning that the reduced density matrix, $\rho$, of a small spatial subsystem of a single, typical, excited energy eigenstate is gaussian with a modular hamiltonian, $h \equiv -\log \rho$, asymptotically proportional to the free fermion hamiltonian. In effect, the ET may be thought of as providing an independent route to establishing the Fermi-Dirac (FD) distribution for free fermions. We revisit this argument for fermions belonging to a superposition of degenerate eigenstates. Specifically, we show that the modular hamiltonian corresponding to a superposition of degenerate eigenstates in a 1-d free fermion lattice system is interacting, and therefore the reduced density matrix for the subsystem is not gaussian. Measurements of the single particle distribution of a small subsystem would then exhibit non-thermal deviations from the FD distribution. These features may be traced to a violation of Wick's theorem for the chosen excited state. We develop a perturbation theory for the modular hamiltonian corresponding to a weak mixture of degenerate states and study the modular hamiltonian (exactly) computationally. Finally, we speculate that initial conditions in a realistic fermi system evolving towards equilibrium might be encoded in the interaction terms of the modular hamiltonian and consequent correlations of fermions.
Read the full abstract
Quantum Physics
Magic Can Superactivate Quantum Capacity
Environmental magic can enable two channels with zero quantum capacity to transmit quantum information jointly. We establish this superactivation within a family of discrete quantum convolution channels with fixed Clifford interactions. For the couplings considered, stabilizer-mixture environments induce entanglement-breaking channels. We construct a mixed magic environment whose induced channel has positive private capacity and a Choi state with positive partial transpose, hence zero quantum capacity. A half-erasure auxiliary activates this channel through the Smith–Yard mechanism. Finite-block encoding and decoding over a self-complementary convolution auxiliary then realize a higher-dimensional erasure channel that still permits activation. We also bound the activated capacity by the modified relative entropy of magic of the main environment. Replacing either environment of the convolution pair by any stabilizer mixture makes the joint quantum capacity vanish, identifying environmental magic as a necessary resource on both sides for this realization.
Read the full abstract
Numerical Analysis
A Hybrid Quantum-Classical Coarse-to-Fine Finite Element Solver for Heterogeneous Darcy Flow
Quantum linear system algorithms offer theoretical potential for solving large sparse systems from PDE discretizations. However, applying them directly to fine-scale problems remains impractical on near-term hardware because of limited qubits, restricted circuit depth, and noise. They also prepare a quantum state rather than solution coefficients, whose recovery requires measurement and reconstruction. In this work, we develop a hybrid quantum-classical coarse-to-fine finite element framework for heterogeneous Darcy flow. We investigate whether a measured quantum coarse solution improves fine-grid initialization and reduces subsequent classical solver work. Rather than solve the full fine-scale system on a quantum device, we apply a method inspired by the Variational Quantum Linear Solver (VQLS) to the Galerkin coarse system. We reconstruct the approximate coarse solution from measurements, prolongate it to the fine grid, and use it to initialize a classical preconditioned iterative solve.} Numerical experiments on steady heterogeneous Darcy flow and a time-dependent SPE10-based problem show that measured quantum coarse solutions can improve fine-grid initialization, while classical correction recovers accurate pressure fields. The effect on classical solver iterations depends on the problem. All quantum computations were simulated using Qiskit Aer.
Read the full abstract
Probability
Discrete total variation under the multinomial, Part 4: variance under a general distribution
We study the variance of the reference-centered discrete total variation $d_{\mathbf p}=\sum_{j=1}^{n-1}|X_{j+1}-X_j-N(p_{j+1}-p_j)|$ of $X\sim\mathrm{Mult}(N;\mathbf p)$ for fixed $n\ge3$ and fixed $\mathbf p$ with all $p_i>0$, as $N\to\infty$. We prove $\mathrm{Var}(d_{\mathbf p})=N\,v(\mathbf p)+O(1)$, where $v(\mathbf p)$, the variance of the sum of absolute values of the limiting Gaussian edge vector, is expressed through the exact multinomial edge variances and correlations and Nabeya's covariance formula for folded bivariate normal variables. Unlike in the uniform case, non-adjacent edge pairs generally contribute to $v(\mathbf p)$. The $O(1)$ remainder is controlled by a lattice Edgeworth expansion and Poisson summation; it oscillates with the lattice phases $\{N(p_{j+1}-p_j)\}$, and for $n=3,4$, where a pair of edges exhausts the bins, the summation runs over a sublattice coset. The formula reduces to the known uniform result (the classical Hoel–Kamat variance), gives a closed form when the reference law is a symmetric comb, and has a shape-independent large-bin limit for positive $C^1$ profiles. For the statistic $d_{\mathbf r}$ centered at a reference law $\mathbf r\ne\mathbf p$ the absolute values linearize: when every edge is mismatched the variance is that of a linear form in the counts up to an exponentially small error, and in mixed (triangular-array) regimes it is given to $o(N)$ by bivariate folded-normal covariances. The two-bin formulas are recorded from the companion preprints, and exact finite-$N$ reductions for $n\le4$ are used for validation.
Read the full abstract
Signal Processing
Multi-Mode Pinching-Antenna-Based OFDM Systems: A Robust Design Perspective
A robust transmission framework is proposed for a multi-mode pinching-antenna (MPA)-based orthogonal frequency-division multiplexing (OFDM) system under both channel state information (CSI) errors and PA positioning errors. A physical model is first developed based on the multi-mode coupled-mode equations (CMEs), which characterizes the frequency-dependent guided-mode propagation and the coupling among guided modes and PAs over the OFDM band. Based on this model, a worst-case sum-rate maximization problem is formulated to jointly optimize the baseband beamformer and the intended PA positions under CSI and PA positioning errors, subject to user quality-of-service (QoS) and transmit-power constraints. To solve the resulting non convex problem, the worst-case CSI is exactly obtained for any given PA positions by solving trust-region subproblems, while the PA positioning errors are handled by grid-based and second-order approximation-based scenario constructions. Then, an alternating optimization (AO) algorithm is developed, where the baseband beamformer and the intended PA positions are updated via the weighted minimum mean-square error (WMMSE) method and successive convex approximation (SCA), respectively. Both subproblems are solved efficiently by exploiting the Karush–Kuhn–Tucker (KKT) conditions. Numerical results show that the proposed robust design outperforms the non-robust and conventional robust benchmarks with the same multi-mode architecture in terms of both worst-case sum rate and QoS satisfaction. They also demonstrate that the proposed robust design satisfies the QoS requirements of all users more reliably than the baselines that consider either CSI errors or PA positioning errors.
Read the full abstract
Analysis of PDEs
Generalized and Classical Harnack Inequalities for semilinear equations on Carnot Groups
We prove generalized and classical Harnack inequalities for non-negative smooth solutions of divergence-form semilinear subelliptic equations on Carnot groups. The proof combines smooth horizontal Gauss–Green and coarea formulas, energy estimates for level sets, a Vitali covering argument, and a horizontal Keller–Osserman barrier. The companion paper, Reference [10], develops the geometric foundations and extends the estimates to weak solutions with nonsmooth data.
Read the full abstract
Data Structures and Algorithms
Dual-Number Associative Arrays for One-Pass Provenance and Sensitivity
Associative arrays and GraphBLAS express sparse computations as matrix operations over a chosen semiring, but the result normally contains only the ordinary value. We define a dual-number extension $D_k(S)=S\ltimes S^k$ that carries this value together with $k$ independent auxiliary channels in the same computation. We prove that the construction is a split square-zero extension of any commutative semiring. Over min-plus, the auxiliary value is the weight of the shortest walk constrained to use the corresponding edge, excluding the weight of that edge. This gives an exact shortest-path membership test for that edge, although ties return the union of the edges on all optimal paths. We implement the min-plus construction using sparse python-graphblas operations, with a fixpoint test that checks the value and auxiliary components together. We validate the implementation across 18,692 checks, including 2,825 tied target pairs. On random graphs with 100,000 vertices and 32 seed edges, the dual method is $3.39\times$ faster than the sequential algorithm but $1.50\times$ slower than a min-plus-specific batched method. This difference is largely explained by the dual method requiring $1.55\times$ as many sweeps. Thus, the main contribution is a general algebraic approach with competitive sparse performance, rather than information that specialized min-plus algorithms cannot already compute.
Read the full abstract
Solar and Stellar Astrophysics
SQUALO II: Probing Accretion Through CO Molecular Outflows in Massive Star-forming Clumps
Molecular outflows probe protostellar accretion and mass transfer from clump scales to forming stars. We present an ALMA study of CO (J=2-1) outflows toward 13 massive clumps from the SQUALO sample, spanning a broad evolutionary range and exhibiting clump-scale infall. We identify 51 high-velocity CO lobes associated with 31 outflow systems and derive their physical and dynamical properties assuming local thermodynamic equilibrium. Outflow inferred accretion rates are typically 10^{-5}-10^{-3} Msun/yr, consistent with active massive star formation. The maximum outflow velocity shows little dependence on candidate driving-core mass, indicating that it is not a reliable tracer of protostellar mass in clustered environments. Outflow momentum flux correlates positively with clump surface density, whereas accretion rate shows a tentative, statistically nonsignificant decrease with increasing virial parameter. Total outflow inferred accretion rate shows a tentative positive relation with clump-scale infall rate but remains systematically lower, suggesting that the detected outflows represent only part of the large-scale mass flux. The clump L/M ratio moderately anticorrelates with projected outflow length, although projection, observational limitations, and the embedded source population prevent a unique evolutionary interpretation. Together with the continuum derived fragmentation properties, our results show that compact cores embedded within infalling clumps drive the outflows. The tentative relation between clump-scale infall and protostellar accretion suggests a possible connection across these scales, although our observations do not directly trace mass transfer from filaments to individual protostars.
Read the full abstract
Mathematical Physics
Finite-strain second-gradient elasticity with an energetic surface: existence of minimizers, incremental theory and elastocapillary beading
We develop a finite-strain theory of second-gradient elasticity for solids with a material surface endowed with its own constitutive energy. The bulk second gradient plays two complementary roles. First, under suitable coercivity assumptions and convexity with respect to the second gradient, we prove the existence of minimizers for nonnegative, lower semicontinuous surface energy densities, without requiring convexity, polyconvexity, positive one-homogeneity, or an upper growth bound on the surface energy. Second, it introduces an intrinsic bulk length scale that affects the onset of instability and wavelength selection. To investigate these effects, we derive an incremental small-on-large formulation in the finitely deformed configuration and apply it to the elastocapillary beading of an axially stretched soft cylinder. For an infinite cylinder and the parameters considered, increasing the second-gradient modulus delays bifurcation and shifts the critical mode towards longer wavelengths. Above a finite threshold of this modulus, the critical stretch saturates in the long-wave limit. The model is implemented in the open-source finite element code OOFEM using C1-continuous isogeometric analysis. Finite-cylinder simulations yield onset stretches in close agreement with the analytical predictions. Selected post-bifurcation branches traced by arc-length continuation exhibit beaded states and supercritical bifurcation at small second-gradient moduli, whereas larger moduli are associated with subcritical bifurcation and the formation of distinct thick and thin segments.
Read the full abstract
Mathematical Physics
Absence of Excited Bound States for All Fibers of the Fröhlich Polaron at Weak Coupling
We study the discrete spectrum of the three-dimensional Fröhlich polaron at weak coupling and fixed total momentum. We prove that, for sufficiently small coupling and every total momentum $0<|\xi|\le 1/4$, the ground-state eigenvalue is the only eigenvalue below the essential spectrum. Our proof builds on the analogous zero-momentum result of Seiringer (2023). Combined with that result and the work of Dahlbaek (2017), this implies that, at sufficiently weak coupling, no fiber Hamiltonian has a discrete excited eigenvalue.
Read the full abstract
Networking
Scalable Full-Horizon SFC Deployment over Predictable SAGIN Topologies via a Configurable Column Generation Framework
Space-air-ground integrated networks (SAGINs) integrate heterogeneous communication and computing resources to support ubiquitous services. Service function chains (SFCs) enable complex services to be composed of ordered functions and executed collaboratively across these resources. Mobility continuously reshapes connectivity throughout service execution, yet much of the resulting topology evolution is predictable. Decisions based only on isolated topology snapshots may therefore overlook valuable future connectivity. We formulate an admission-first, cost-aware model on a time-expanded graph for SFC deployment. Our key observation is that temporal feasibility is request-specific, whereas global coupling arises mainly from shared-resource competition. We therefore develop a configurable column-generation framework in which complete request deployments form columns and a restricted master problem coordinates admission and resources. SAGINFE, a fast constructive heuristic built on per-flow embedding (PFE), supplies a standalone feasible solution and initial columns. Configurable pricing schedules combine early-terminated and stricter searches, supplemented by selective PFE candidates. Candidates undergo feasibility and exact reduced-cost checks. Request-wise pricing can run in parallel, followed by finite-pool integer selection. The default configuration improves the admission-first objective over initialization on all 120 main instances. Four pricing workers achieve up to a 2.69-fold median pricing-stage speedup while preserving admitted request counts in the tested cases.
Read the full abstract
Robotics
Extended Model-Based Learned Inertial Odometry
Inertial odometry is a compelling approach to state estimation for agile quadrotor flight due to its affordability, low weight, and robustness to perceptual degradation. However, relying only on integrated inertial measurements is impractical, as sensor errors and time-varying biases lead to significant pose drift. Although recent advances have enabled inertial odometry for drone racing, existing approaches show limited generalization to trajectories not seen during training. This work improves generalization by adopting a body-frame representation and incorporating body-frame torque dynamics into the learning pipeline. The learned model predicts short-horizon relative displacements, which are fused with IMU measurements in an Extended Kalman Filter. Experimental results show that the proposed method outperforms the previous state-of-the-art learned inertial odometry approach for quadrotor pose estimation on unseen trajectories. An additional evaluation on four public quadrotor datasets, with whole trajectory shapes held out of training, confirms a lower velocity error on every held-out set and shows that the torque input is most informative when it is derived from the rotor signals. Code, data pipeline and trained models are publicly available.
Read the full abstract
Computer Vision
MSR-ABC: Complementary Feature Fusion for Clear Cell Renal Cell Carcinoma Classification
Clear cell renal cell carcinoma (ccRCC) classification from computed tomography (CT) may benefit from representations that capture different aspects of tumour phenotype. However, simply concatenating reconstruction-derived features, image–text embeddings, and radiomic descriptors may not fully exploit their complementary information. We propose MSR-ABC, a global–local framework integrating MedVAE, MedSigLIP, and radiomics. Spatially matched local image descriptors interact through cross-representation attention, while radiomic auxiliary supervision encourages the pooled local representation to retain quantitative tumour attributes. A gated residual mechanism integrates local information with global summaries. Area-aware pooling and visual and soft-text prompting complete the framework. We evaluate discrimination using a patient-level split of 396 KiTS23 cases and an external cohort of 111 TCGA/TCIA cases, reporting mean AUROC over three training seeds with 95% patient-bootstrap confidence intervals. MSR-ABC achieves mean AUROCs of 0.806 internally and 0.879 externally. Internally, it exceeds global MedVAE–radiomics and three-source concatenation by 3.35 and 3.61 AUROC percentage points, respectively. Removing any source lowers mean AUROC in both cohorts, with the largest decreases following radiomics removal. Individual removal of cross-representation attention, auxiliary supervision, or gated fusion also lowers internal mean AUROC. These findings are consistent with complementary contributions within the integrated architecture and support further investigation of structured representation fusion for non-invasive renal tumour characterisation.
Read the full abstract
Nuclear Theory
Mathematical foundations and numerical simulations of BDNK hydrodynamics: from heavy-ion collisions to astrophysics
We present new developments in the mathematical foundations and numerical simulations of the causal and stable first-order relativistic viscous hydrodynamic theory by Bemfica, Disconzi, Noronha, and Kovtun (BDNK) in general spacetimes. We develop two independent flux-conservative algorithms for arbitrary 3+1-dimensional spacetime metrics. Both are hyperbolic, causal, and well-posed, pass convergence tests, and reproduce known results. The first uses a new first-order reduction with mathematically preserved constraints, whose numerical violations can be handled with cleaning fields (the resulting coupled system is proven hyperbolic and well-posed). The second uses a new mixed-order formulation, whose solutions are proven strictly equivalent to the original BDNK system, without cleaning fields. Our proofs rely on a new general theorem providing sufficient conditions for transforming a mixed-order system with arbitrary-order equations into a strictly hyperbolic system, strengthening a classical result by Choquet Bruhat for weak hyperbolicity. We introduce a new practical prescription for selecting admissible hydrodynamic frames for a broad class of equations of state, including nonconformal ones relevant to heavy-ion collisions and astrophysics. Our simulations capture viscous shock broadening and suppression of Kelvin–Helmholtz vortices, reproduce Bjorken and Gubser flows, and study departures from stationary ideal-fluid configurations in Michel accretion and rotating tori on fixed Schwarzschild and Kerr backgrounds. For heavy-ion collisions, we perform boost invariant simulations with a lattice-QCD-based equation of state and fluctuating TRENTo initial conditions, comparing BDNK with a minimal Israel–Stewart theory. The descriptions agree at small shear viscosity but can differ substantially at larger viscosities, particularly for early initialization with rapid longitudinal expansion.
Read the full abstract
Mesoscale and Nanoscale Physics
Scanning Probe Control, Quantum Dynamics, and Dispersive Readout of Suspended Carbon Nanotube Resonators
Suspended carbon nanotubes (CNTs) offer promising opportunities for quantum nanomechanics owing to their small effective mass, strong electrostatic tunability, and mechanical nonlinearity. We develop a unified theoretical framework for coherent control and readout of a suspended CNT driven by an atomic force microscope (AFM) tip and capacitively coupled to a Cooper-pair box (CPB). We first derive the linear and nonlinear Duffing response of the CNT and quantify the effects of AFM force gradients, tip-position fluctuations, and voltage noise on mechanical stability, dephasing, and heating. These results establish operating conditions for coherent actuation with limited backaction. We then quantize the Duffing oscillator, derive its Kerr anharmonicity, and identify the requirements for selective control of individual vibrational transitions. Estimates of thermal occupation and active cooling illustrate the constraints for CNT modes spanning frequencies from 54 MHz to 5.37 GHz. Finally, we derive the CNT–CPB interaction and its dispersive limit, showing how phonon number and parity can be mapped onto the CPB through Ramsey interferometry and detected using quantum-capacitance readout. The resulting framework identifies the conditions under which localized AFM actuation, nonlinear quantum dynamics, and superconducting state-resolved detection can operate within a single CNT device.
Read the full abstract
History and Overview of Mathematics
Hostages
Our house has been taken over. It is not completely destroyed, but many of its essential functions and facilities no longer work as they used to. Essentially, they are out of our control. Artificial intelligence, through its large language models (LLMs), has taken it. And we, the inhabitants of the house, without really noticing what was happening, and surprised by the effectiveness of LLMs, have handed over our most sacred belongings: the capacity to think about a problem, the joy of having an idea that solves it, the beautiful effort needed to consolidate an idea in our common language. And there is no turning back. We are mathematicians, and our whole world is in check. We wonder: is it checkmate? Can we take back control of our house, which will undoubtedly not work in the same way? Paradoxically, the weapons that attacked us are built on some of our own ideas. Of course, it is not entirely our work. The computer science knowledge needed to develop these large models evolved in both its theoretical and its material aspects. As technology evolved (infinitely small processors, infinitely large computing centers), turning quantity into quality, computer scientists encoded into software the ideas needed to produce these tools (let us not call them weapons). These notes look into the past and the future of the mathematical profession. What will inevitably happen? What will remain and what will disappear? And, in consequence, what is our line of action? To organize the ideas, after some introductory historical comments, we try to answer the questions: (1) Is mathematics necessary for the development of our society? (2) Will mathematicians be necessary in the near future? (3) How the profession will change, through a list of exercises: open questions and possible actions.
Read the full abstract
Statistical Computation
Locally Adaptive Conformal Prediction for Polynomial Chaos Expansions
This work develops a locally adaptive jackknife+ conformal prediction framework for polynomial chaos expansions. Therein, the non-conformity scores are normalized with an auxiliary polynomial chaos expansion trained on the log-transformed leave-one-out residuals of the base model. We show that the coverage guarantee of jackknife+ conformal prediction carries over to the adaptive intervals if the leave-one-out auxiliary models are trained on nested leave-one-out residuals. For fixed-basis polynomial chaos expansions fitted by ordinary least squares, all required quantities are derived in closed form and the guarantee-preserving adaptive intervals are obtained from a single model fit, without any model retraining. For sparse polynomial chaos expansions, where multiple model fits are necessary, we propose a heuristic simplification that limits the computational overhead of adaptivity at the cost of the formal coverage guarantee. Numerical experiments on an academic benchmark with known heteroscedastic noise and on an engineering simulation dataset show that the suggested method attains the target coverage while providing intervals that adapt to local prediction uncertainty, therefore being considerably more statistically efficient. These advantages are particularly prominent in high-uncertainty regions of the input space, where the performance of non-adaptive methods degrades substantially.
Read the full abstract
Mesoscale and Nanoscale Physics
Quantum Interference Control of Coupled Atomic Spins
We propose a protocol for the non-resonant electric control of quantum states in atomic-scale spin systems using Landau-Zener-Stückelberg-Majorana (LZSM) interferometry. We focus on a Ti-Ti-Fe trimer on a thin MgO layer, a platform that has recently emerged as a promising candidate for qubit realization using electron spin resonance combined with scanning tunneling microscopy (ESR-STM). We first analyze a Ti dimer in the absence of the Fe atom and show, through a detailed spectroscopic study, that avoided level crossings can be engineered by combining the in-plane component of the external magnetic field, the tip magnetic field, and the applied bias voltage, which allows the system to be tuned electrically through these crossings. Based on this, we develop LZSM protocols for coherent state control in the dimer using non-resonant LZSM-STM. We then consider the full trimer, where the Fe atom generates a local magnetic field at the remote Ti atom, enabling non-local spin control. We show that this remote and non-resonant manipulation reduces unwanted effects associated with tunneling electrons and enables quantum operations on short timescales. Our results show that quantum interference can be used as a resource for electrically driven, non-resonant, and non-local spin control in atomic-scale systems. This approach enables fast quantum operations while mitigating decoherence, offering a promising route toward scalable atom-based qubit architectures.
Read the full abstract
Functional Analysis
Isometric Classification of Sobolev Spaces
We show that the norm obtained by summing the $L^1$ norms of all weak derivatives through order $k$, including the function itself, determines both $n$ and $k$ on the real Sobolev space $W^{k,1}(\R^n)$. Every surjective linear isometry for this norm is given by a translation, a signed permutation of the coordinates and multiplication by a fixed sign $c\in\{-1,1\}$.
Read the full abstract
Signal Processing
Expert Spindles in the Wild: Enriched Supervision for a Dense Detector
Sleep spindles are a hallmark of healthy NREM sleep whose frequency, amplitude, density, duration and topography vary with brain maturation, age and disease. Their manual annotation is impractical and variable between scorers, yet the public references used to train automatic detectors cover these characteristics only partially: stage N2 only, a single central derivation, or short pre-selected segments. Measuring duration and density requires dense detectors that learn event boundaries from annotated onsets and offsets, and therefore inherit the protocol under which those boundaries were scored. We introduce Expert Spindles in the Wild (ESW), a stage-agnostic expert spindle annotation set covering 19 complete MASS-SS2 recordings ($\approx$150 h of continuous EEG, 29,819 events). A sleep neurophysiologist annotated the recordings using the full EEG montage, blinded to sleep-stage labels and prior spindle annotations. On the 15 nights shared with historical scorers, ESW recovers 92.7% of E1 and 73.7% of E2 events, while recovering 81.6% of the MODA expert-consensus events on the common excerpts. Despite stage-blind annotation, 99.0% of ESW events occur in N2/N3. Used to train a fixed dense detector on a frozen SleepFM encoder, ESW reaches an event-level $F_1$ of 0.680 on 166 unseen MODA subjects, compared with 0.635 for E2 ($\Delta F_1$ = +0.045, paired-bootstrap 95% CI [0.024, 0.065]). ESW is thus useful supervision for cross-dataset detection, and closer annotation-level agreement with the MODA consensus did not translate into better transfer performance in this experiment. The annotations are released under the MASS data-use terms (https://doi.org/10.5281/zenodo.23894975).