Machine Learning
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Wednesday 9 September
Simultaneous assessment of medical imaging and patient records is often required in clinical diagnosis. However, standard machine learning algorithms cannot analyze these data types together. Meanwhile, compliance with HIPAA and GDPR can constrain centralized aggregation of sensitive patient data. This leaves a crucial void of secure fusion of visual and textual context across distant networks. Thus, we present OmniMed-FL, a controlled systems study of multimodal federated learning for five-class clinical condition classification (Normal, Pneumonia, COVID-19, Pleural Effusion, Cardiomegaly). Our proxy corpus pairs 3,000 public chest radiographs with 3,000 class-conditioned synthetic notes, matched by class, not by patient. The framework benchmarks eight fusion strategies, three initializations, four missing-text imputation rules, and matched federated baselines under non-IID Dirichlet partitioning across 3 to 20 hospital clients. As all notes are synthetic and pairing is not patient-level, these are descriptive proxy comparisons, not estimates of diagnostic performance or deployment readiness. Within those limits with clients ($K=5$) and severe skew ($\alpha=0.1$), local-only training achieves a macro-F1 score of 0.297, FedAvg achieves $0.662\pm0.074$, FedProx $0.737\pm0.085$, a matched FedMME-style one-shot ensemble $0.647\pm0.080$, and our SCAFFOLD-AdamW adaptation $0.070\pm0.015$, the 0.075 FedProx-FedAvg gap falling inside the wider of the two two-seed standard deviations. Over a $4\times3$ grid, label skew costs up to 0.27 F1 whereas a near-sevenfold client increase costs at most 0.10, while bidirectional volume grows linearly to 183.5 GiB at $K=20$. Multimodal fusion leads on both corpora, scoring 0.956 against 0.934 for text and 0.664 for images on the synthetic corpus and 0.906 against 0.880 and 0.737 on the radiograph corpus, for $2.3\times$ the model state of text alone.
Computer Vision
SceneHI: High-Resolution 3D-Consistent Scene Texturing with Controllable Illumination
SceneHI is a framework that lifts high-resolution, illumination-aware priors from 2D diffusion models to perform 3D texture synthesis. It is the first to demonstrate that high-resolution textures, previously limited to 2D synthesis, can be generated directly on 3D objects without model fine-tuning or optimization. Designed for complex, multi-object environments, SceneHI uniquely combines 3D-consistency, high-resolution fidelity, and physically plausible baked shadows within a single generative pipeline. To enforce strict geometric coherence, we introduce an exact analytical pixel-to-texel mapping that aligns diffusion trajectories across multiple viewpoints. We utilize High-Resolution Latent Textures (HRLTs) as a persistent canvas for gradually denoised textures, while camera views perform the denoising steps in latent pixel space. This ensures a shared base texture that can be subsequently refined to high resolution without compromising multi-view consistency. Finally, a light-aware generative pass embeds realistic geometry-consistent shadows directly into the atlases, bridging the gap to production workflows. SceneHI achieves high visual fidelity while reducing generation time by 80% compared to existing scene-level methods.
Read the full abstract
Rings and Algebras
Weak central polynomials for algebras of multiplications of simple algebras
We give a simple proof for the existence of weak central polynomials for the algebra of multiplications of a finite-dimensional simple (non-associative) algebra. As an example we present explicit weak central polynomials in the cases of the three-dimensional simple Lie algebra and the Jordan algebra of the two-dimensional vector space with non-degenerate symmetric bilinear form.
Read the full abstract
Complex Variables
Convexity of the Berezin range of composition operators induced by automorphisms of the Unit Ball in $\mathbb{C}^n$
We study the convexity of the Bergman space Berezin range of composition operators induced by unit disk and unit ball automorphisms. We obtain several new results characterizing convexity of the Berezin range for linear fractional unit disk and unit ball automorphisms composed with unitary maps.
Read the full abstract
Solar and Stellar Astrophysics
On the 3D Magnetic Structure of Coronal Mass Ejections Through the Solar Corona
The magnetic fields that make up the internal structure of coronal mass ejections (CMEs) are thought to be organised in a flux rope configuration, consisting of twisted magnetic fields that wind about a central axis. Remote-sensing observations of CMEs show a wide range of morphologies, dynamics, and evolution, including rotation, non-uniform expansion, deflection, and interaction with the ambient solar wind. In-situ measurements, however, typically consist of a single 1D spacecraft trajectory through a large 3D structure (or few at best), restricting our ability to determine how CME magnetic structure varies in space and time. In this work, we analyse the magnetic configuration of two idealised CMEs during their early evolution in the range 1-30 $R_{\odot}$ using the Magnetohydrodynamic Algorithm Outside a Sphere code. The initial flux ropes differ only in their orientation with respect to the global dipole and erupt from a bipolar active region below the streamer belt in a simplified coronal environment. We deploy a fleet of synthetic spacecraft throughout the CMEs' paths at different combinations of latitudes, longitudes, and heliocentric distances. We identify and examine flux rope signatures in the synthetic in-situ profiles to characterise radial variations as well as latitudinal/longitudinal ones. We find that, even in a simplified CME erupting under solar minimum-like conditions, the sampling location significantly affects the global structure that would be deduced from common flux rope reconstruction and analysis techniques used for in-situ measurements. Additionally, we consider possible in-situ signatures that may serve as indicators of the source region's topology.
Read the full abstract
Statistical Mechanics
Quantifying translational and bond-orientational order metrics in hyperuniform and nonhyperuniform many-particle systems
Quantifying the degree of order/disorder in many-particle systems remains an outstanding problem in physics, materials science, and mathematics. To this end, we consider the translational order metric $\tau_T$, defined as the squared $L^2$ norm of the total correlation function $h(\mathbf{r})$, and introduce its bond-orientational analogue $\tau_O$, defined from the weighted total correlation function $h_{\mathbf{f}}(\mathbf{r})$ using local orientational weights (S. Torquato et al., Phys. Rev. X 16, 011042 (2026)). The pair $(\tau_T,\tau_O)$ places both forms of order on a common two-point statistical footing. We compute both metrics for two nonhyperuniform sphere-packing models in 2D and 3D as functions of packing fraction $\phi$: 1) equilibrium hard particles and 2) nonequilibrium random sequential addition (RSA) packings. For the nonhyperuniform systems, bond-orientational order remains subdominant along the equilibrium-fluid and RSA configurations; however, its magnitude relative to the translational metric increases near the upper end of the equilibrium-fluid branches, much more strongly in 2D than in 3D, and the two metrics become comparable along the sampled crystal branches. At common packing fractions, equilibrium fluids and RSA packings trace distinct $(\tau_T,\tau_O)$ trajectories, revealing preparation-dependent differences in structural order. As a representative hyperuniform family, we study 2D disordered stealthy hyperuniform (SHU) ground states for $0<\chi<1/2$, where $\chi$ is the stealthiness parameter. Within the disordered SHU phase, bond-orientational order remains subdominant, but its magnitude relative to the translational metric increases toward the disorder-to-order threshold. In all three models, $\tau_T$ and $\tau_O$ are positively correlated beyond the Poisson-reference regime: $\tau_O$ increases monotonically with $\tau_T$ across the sampled state points.
Read the full abstract
Probability
Periodic ASEP and random walks on affine Hecke Algebra
We study the large time behavior of a periodic ASEP with fixed period started from the step initial condition. We establish the limit shape theorem and the asymptotic distribution of a single second class particle. Our single-species results extend the previous results of Lam and Ayyer-Linusson about periodic TASEP to the ASEP case. These results can also be interpreted as describing the limit behavior of a random walk on affine Hecke algebra.
Read the full abstract
Machine Learning
A Later Test Set Is Not a New Domain: Pretraining Familiarity Survives a Contamination-Free Hold-Out
Time-series foundation models are evaluated almost exclusively on public archives that predate them, so a strong score cannot be separated from having seen the test set during pretraining. The obvious remedy is a hold-out that postdates the models. We build one: thirteen forecasters – four classical, three trained per dataset, six pretrained – on seven groups drawn from five domains, every observation published after the last model was released, and every dataset rebuildable without an API key. Under this protocol pretrained models win 5 of 7 groups, lose one to a Theta baseline, and on daily exchange rates are indistinguishable from a seasonal naive forecast, along with every other method tested. We then ask what separates the wins from the losses, and report a negative result: the two intrinsic properties one would reach for – seasonal strength and spectral entropy, measured on the input window – do not account for the pattern, and seasonal strength is if anything negatively associated with the advantage. What does track it is corpus familiarity. Our largest gain (28% lower MASE than the best classical method, on weekly Wikipedia pageviews) falls on Wikipedia pageviews, the domain TimesFM's authors describe as the bulk of its pretraining corpus, at the same granularities and differing only in time window. Within the pretrained family, where every model forecasts identical series so that series difficulty cancels, the TimesFM family outranks the Chronos family by -0.53 ranks on Wikipedia against -0.09 everywhere else (1,500 vs. 754 series, Mann-Whitney p < 1e-5). We conclude that a temporal hold-out removes memorisation of a window but not familiarity with a domain, that benchmarks therefore need domain hold-outs stated relative to disclosed corpora, and that the practitioner's question is less which model is better than whether their domain is one the model was raised on.
Read the full abstract
Computer Vision
Spot-the-shift: Evaluating Grounded Image Difference Captioning of Long-term Changes
Long-term change understanding from images of the same place revisited over time is a challenging task with applications in map maintenance and urban infrastructure monitoring. Prior work addresses it either through pixel-level prediction or difference captioning, neither of which is sufficient to reliably measure how well models detect and describe such changes. We introduce SPOT-THE-SHIFT, a human-verified benchmark for grounded image difference captioning of long-term changes in real-world driving scenes. Our benchmark provides natural language captions and spatial masks for structural changes across each image pair. We further propose an evaluation protocol that reliably assesses models' captioning ability, validated through human studies. Benchmarking state-of-the-art MLLMs, we find that models struggle with the fine-grained multi-image spatial capability required for this task. Finally, we develop a synthetic data generation pipeline that improves an off-the-shelf MLLM without sacrificing general capabilities.
Read the full abstract
Computer Vision
Why Is Video Still So Expensive? A Survey of Inference-Efficiency Mechanisms in Video and Audiovisual LLMs
Video understanding has rapidly evolved toward video large language models (VideoLLMs): systems that couple video representations with pretrained large language models and condition generation on a textual prompt. Their strong performance on captioning, question answering, retrieval and temporal grounding comes at a computation and memory cost that grows with frame count and context length, limiting deployment in real-time, mobile and resource-constrained settings. This survey covers inference-efficiency mechanisms for visual and audiovisual VideoLLMs that report concrete reductions in parameter count, FLOPs per input, latency, memory, or visual and audio token count. We analyze bottlenecks across frame sampling, modality encoding, connector-level token reduction, and LLM prefilling and decoding. We organize methods by the pipeline stage at which they act, covering VideoLLMs developed since late 2022 together with earlier frame-sampling and vision-encoder mechanisms that remain components of current pipelines. We assemble literature-reported accuracy–cost comparisons under shared host models and input protocols wherever available, distinguish them from heterogeneous cross-paper evidence, and identify gaps in audiovisual efficiency and standardized evaluation. We maintain a repository at https://github.com/momentslab/awesome-efficient-videollm.
Read the full abstract
General Relativity and Quantum Cosmology
A Novel Scheme for Inter-Satellite Integrated Laser Ranging and Communication in Space-Based GW Detection
Adopting pseudo-random noise (PRN) codes for inter-satellite ranging, the resulting data-encoded PRN signal induces excessive laser phase measurement noise under the currently adopted integrated scheme in space-based gravitational wave (GW) detection systems. To suppress the noise, a scheme of inter-satellite integrated laser link is presented in this Letter, in which the data-encoded PRN signal for inter-satellite absolute distance measurement and communication is phase-modulated onto the ultra-stable oscillator (USO) clock signal before phase modulation onto the laser carrier. Theoretical analysis and simulation results show that in the proposed scheme the noise is significantly suppressed compared with the currently-adopted one, while the noise stemming from the data-encoded PRN signal on clock noise transfer is limited within the requirement.
Read the full abstract
Algebraic Geometry
Birational cubic fourfolds via an Enriques Cremona transformation
We describe a Cremona transformation on $\mathbb{P}^5$ given by the linear system of quintics that are double along the Fano model of an Enriques surface $S$. Using this Cremona transform, we construct a birational equivalence between the very general cubic fourfold $X$ of discriminant 44 and its unique Fourier–Mukai partner. We also classify birational maps between smooth, non-isomorphic cubic fourfolds resolved by blowing up a smooth surface on each side: the centres must be either Fano models of Enriques surfaces, as in our construction, or Veronese surfaces, as studied by Fan–Lai.
Read the full abstract
Mathematical Physics
Universal correlations in the Abelian sandpile model
We numerically study the bulk correlation functions in the two-dimensional Abelian sandpile model, aiming both to compare with the predictions of logarithmic conformal field theory and to extend the analysis to lattices where analytical methods are difficult to apply. Wilson's algorithm efficiently generates large-scale uniform spanning trees in parallel, which can be mapped to independent recurrent configurations via the Majumdar–Dhar burning bijection, eliminating sample autocorrelations and yielding fast convergence. On the square (single-sublattice) and honeycomb (two-sublattice) lattices our results agree well with the analytical predictions. For the kagome lattice we provide the first systematic numerical study of the bulk correlation functions, and cross-check the bulk height-1 probability against a closed-form analytical expression that we also derive here via the lattice Green function.
Read the full abstract
Sound
Unifying Score and Performance for Fine-Grained Music Understanding in Audio-Language Models
Large audio language models (LALMs) have shown promising progress in broad music-understanding tasks such as tagging, retrieval, and captioning. Music understanding that requires finer hearing over both the content and how it is realized within a performance through dynamics, phrasing, articulation, time, and other performance techniques, however, remains at an earlier stage. Existing audio-language model (ALM) training pipelines typically rely on coarse, weakly grounded captions and therefore provide little support for learning these subtle nuances in music, limiting their ability to serve real-world applications in education or artistic practice. We therefore introduce MuNo-SP (Music Notation unifying Score and Performance), a text-based representation that jointly encodes score content and performance information. Building on MuNo-SP, we develop an automatic training-data generation pipeline that uses aligned scores and performances to produce long-form auditory analyses and musically informed question-answer pairs. We use this pipeline to construct MAESTROCaps, a classical piano dataset comprising 148 long-form performance analyses and 31,080 question-answer pairs derived from 148 aligned score-performance pairs. In a human evaluation, MuNo-SP analyses were preferred by majority vote over MIDI-only analyses for eight of nine excerpts. MuNo-SP also performed strongly on a benchmark of score-performance understanding, suggesting that integrating score and performance information enables more reliable and musically informative LALM supervision than a MIDI-only baseline.
Read the full abstract
Artificial Intelligence
Cyber-Financial Contagion: Modeling the Propagation of an AI Vendor Compromise Through the Banking System
The banking system now depends on a small set of shared artificial intelligence vendors for fraud screening, credit decisioning, anti-money-laundering triage, customer analytics, and internal decision support. This paper studies how a compromise inside one of those vendors can propagate along a chain of operational, informational, and financial linkages until it triggers losses that look, from the outside, like a classical banking crisis. We build a four-layer heterogeneous network that couples AI vendors, financial institutions, interbank exposures, and customer accounts, and we propose CFC-Prop, a stochastic epidemic-and-clearing model that runs on that network. On a synthetic dataset with 60 vendors, 220 banks, roughly 2,500 vendor-bank service edges, and 1,400 interbank exposures, CFC-Prop reproduces the heavy-tailed loss distributions and the sharp dependence on patch latency that are consistent with prior cyber-financial evidence. We also train an early-warning model, CFC-GNN, that uses vendor-side incident telemetry and graph structure to flag high-cascade-risk vendors before impact. Across four baselines the proposed model reaches AUROC 0.82 and AUPRC 0.60 while keeping calibration errors bounded. We release the full code, synthetic data, and reproducible scripts. The results argue that cyber concentration among AI vendors is a first-order financial-stability problem and give supervisors a concrete quantitative tool for reasoning about it.
Read the full abstract
Operator Algebras
The Cuntz-Pimsner algebra of the simplest Motzkin subproduct system is 2-subhomogeneous
The Motzkin subproduct system (SPS) is constructed from the Jones Wenzl idempotents of the Motzkin algebras, which generalizes the Temperley Lieb SPS. The simplest Motzkin SPS, which is not a Temperley Lieb SPS, is constructed from a 3 dimensional Hilbert space. We explicitly describe the spectrum of the corresponding Cuntz Pimsner algebra, which remarkably admits only irreducible representations of dimensions 1 and 2, and can thus be viewed as a mildly quantum space. We moreover analyse representations of the Richard Thompson groups and of the Cuntz algebra that are associated to this spectrum.
Read the full abstract
Materials Science
Spectroscopic signatures of mode-selective electron-phonon coupling in transient reflectivity change on charge-density-wave TiSe$_2$
We report on time- and spectrally resolved pump-probe spectroscopy measurement on the charge-density wave (CDW) state of TiSe$_2$ at helium temperatures. The non-oscillatory component of the spectrally resolved signals exhibit a sign reversal near 1.5 eV, which was interpreted as a redshift of the optical transition energy originated from the interband transition between the Se $p$-based valence band to the Ti $d$-derived conduction band. Furthermore, the oscillatory components exhibit distinct spectral dependences between the optical phonons and the CDW-derived modes, suggesting the mode-selective coupling.
Read the full abstract
Hardware Architecture
CertiFlash: A Formal Verification Framework for Flash Translation Layers in Computational Solid State Drives
Data-intensive applications move large amounts of data from storage to the compute unit, incurring significant data movement overhead. Storage-centric computing reduces this overhead by moving computation near or inside solid-state drives (SSDs). Enabling it requires modifying SSD policies, e.g., address translation and garbage collection, which are part of the Flash Translation Layer (FTL), the SSD's firmware. Modifying the FTL is error-prone. Because FTL logic has direct access to security-critical device components, even a functionally correct FTL can leak data between tenants, drop integrity tags, or assign a flash block to the wrong tenant. We show that a faulty FTL can corrupt the device state at five surfaces inside the SSD, and demonstrate them on a DaisyPlus OpenSSD. Prior work verifies individual FTL designs, but has two limitations. (1) It establishes only functional correctness, so a modified FTL can violate isolation, integrity, and ownership and still pass verification. (2) It is tied to a single FTL design, so every modification requires redoing every proof. We propose CertiFlash, a formal verification framework for FTLs, mechanized in the Rocq proof assistant, that gives designers a machine-checked proof of security and correctness. CertiFlash models an FTL as a deterministic state machine with a single global invariant over mapping, isolation, integrity, ownership, and allocation. We prove once, over a general FTL model, that (i) every FTL operation preserves the invariant and (ii) the model refines an idealized block device. For a new design, a designer discharges five hypotheses about its own operations instead of redoing either proof. Across four case studies, a designer adds 27 to 3,231 lines against a 16,489-line framework, significantly reducing the verification effort. CertiFlash is open source.
Read the full abstract
Computer Vision
Beyond One-Size-Fits-All: Sample-Adaptive Strategy Routing for Vision Token Pruning in MLLMs
Multimodal large language models (MLLMs) process hundreds or thousands of visual tokens per image, incurring prohibitive inference costs. While existing vision token pruning methods mitigate this overhead, they implicitly assume that a single fixed pruning strategy can be applied uniformly across all inputs. Our analysis further reveals that ranking pruning methods by average benchmark accuracy conceals substantial sample-wise complementarity: although the average-best strategy excels overall, alternative strategies prove superior on a significant fraction of individual samples. To harness this diversity, we propose VIP-Router, a lightweight VIsion Pruning Router that adaptively selects the pruning strategy predicted to be best suited to each input at a specified pruning level. Conditioned on low-cost visual and textual features, VIP-Router identifies the most suitable candidate strategy while retaining full-token inference as an option when pruning is predicted to be unfavorable. Evaluated on a curated suite of pruning-sensitive visual perception benchmarks, VTC-Bench Group A, VIP-Router consistently outperforms the best fixed strategy baseline across all reduction ratios, achieving a 26.9% relative improvement in average accuracy, and a 22.0% relative increase in average utility after accounting for realized token cost. Crucially, VIP-Router operates in a plug-and-play manner without modifying underlying pruning algorithms or model weights, introducing trainable parameters equivalent to merely 0.017% of the backbone. Furthermore, VIP-Router proves effective across various MLLM backbones and yields consistent gains on unseen benchmarks, highlighting the potential of sample adaptive routing for visual token pruning.
Read the full abstract
General Relativity and Quantum Cosmology
Parity symmetry as a diagnostic for spin-precessing binary-black-hole waveform models
Parity invariance of general relativity imposes exact relations between waveforms and remnant properties of binary-black-hole configurations related by reversal of the in-plane spins and observer direction. We summarize these relations and construct consistency tests for spin-precessing waveform and remnant models. Applying a conservative, unmaximized overlap diagnostic to seven state-of-the-art waveform models, we find that models built from aligned-spin calibrated waveforms and parity-covariant precession dynamics satisfy the symmetry to numerical precision. The original IMRPhenomXO4a and IMRPhenomXPNR implementations instead exhibit median parity residuals of approximately $1.6\times10^{-3}$ with broad upper tails, while NRSur7dq4 shows its largest violations at high mass ratio and large spin magnitude. The asymmetric extension of SEOBNRv5PHM exhibits a smaller, localized violation, while NRSur7dq4Remnant shows larger residuals in the recoil than in the final mass and spin. We trace these violations to an observer-dependent phase anchor in the IMRPhenom antisymmetric mode, non-invariant parameterization of the NRSur7dq4 surrogate fits, and one non-invariant NR-calibrated coefficient in the asymmetric SEOBNRv5PHM extension. Local corrections restore waveform parity to numerical precision for the IMRPhenom and SEOBNRv5PHM cases; restoring it in NRSur7dq4 requires refitting the surrogate. Reanalyses of eight GWTC-5.0 events show no qualitative change in astrophysical interpretation, although the IMRPhenom correction yields Jensen-Shannon divergences up to $5\times10^{-2}$ in individual marginalized posteriors. Exact symmetries thus provide inexpensive, model-independent consistency tests and should be imposed by construction in calibrated waveform models.
Read the full abstract
Algebraic Geometry
Categorical reconstruction of del Pezzo surfaces: Hochschild–Serre algebras and spinor modifications
We prove that, for every smooth complex del Pezzo surface of degree at most four, the enhanced right orthogonal to the structure sheaf determines the surface up to isomorphism. In degrees one, two, and three, we recover the anticanonical equation from intrinsic pieces of the Hochschild-Serre algebra via graded matrix factorizations; in degree four, the relevant Serre diagonal recovers the orbifold canonical ring associated with the pencil of quadrics. We also give an alternative proof in degrees one and two using the Bertini and Geiser involutions, equivariant topological K-theory, and classical Torelli. Then, we study the Clifford component associated with a conic bundle structure over the projective line. In every degree at most four, we construct an abstract spinor bundle whose spinor modification produces another, generally non-isomorphic, del Pezzo surface. We show via Kuznetsov's modification theorem that the relevant base-linear equivalences are precisely those induced by spinor modifications.
Read the full abstract
Group Theory
Sharp bounds for intervals in finite subgroup lattices
Let H be a subgroup of a finite group G, and put n = [G:H] > 1. If p is the least prime divisor of n, we prove that the number of subgroups K with H <= K <= G is less than c(p) n^((log_p n)/4). Here c(p) is the product of (1 - p^(-j))^(-1) over all positive integers j, multiplied by the sum of p^(-z^2) over all integers z. The exponent and the constant are simultaneously optimal, already for elementary abelian p-groups of even rank. In particular, for arbitrary n > 1, the number of intermediate subgroups is less than 7.371968802 n^((log_2 n)/4). This removes the polynomial factor from the best previously known relative bound and extends the sharp absolute subgroup-counting theorem to arbitrary subgroup intervals. The group-theoretic argument is elementary. Its main ingredient is a weighted packing inequality for relative generating tuples, obtained from an escape tree on the coset space G/H.
Read the full abstract
Mesoscale and Nanoscale Physics
Quench dynamics in nonreciprocal Aubry-André-Harper model
The critical phase of a non-Hermitian quasicrystal can support stronger transport than its surrounding delocalized phase. We demonstrate this anomalous behavior in the one-dimensional nonreciprocal Aubry-André-Harper model through a combined study of dynamical quantum phase transitions (DQPTs) and wavepacket diffusion. Using a parity-sorted energy-spectrum classification that directly encodes the generalized $\mathcal{PT}$ symmetry, we find that DQPTs in this system are energy-resolved, in contrast to the energy-independent DQPTs of Hermitian quasicrystals. The energy-resolved features are most pronounced when the initial and final Hamiltonians belong to different phases (localized or extended), and they are tied to the even-odd index structure of the spectrum, which we exploit to organize the quench-dynamical landscape. For wavepacket dynamics after a single-site quench, the diffusion exponent $\beta$, extracted from the long-time power-law scaling of the root-mean-square displacement $\sigma(\tau)$, partitions the phase diagram into four distinct regimes. In the Hermitian limit the extended phase is ballistic ($\beta=1$), the critical phase is normally diffusive ($\beta=0.5$), and the localized phase yields $\beta\to 0$. Nonreciprocity reverses this hierarchy: the extended phase becomes normally diffusive, while the critical phase turns ballistic. We trace the anomalous $\beta=1$ at criticality to the self-similar multifractal structure of the critical eigenstates, whose nodal positions are organized by the golden ratio. A finite-size scaling ansatz built on the wave-front propagation yields $\sigma(\tau)\propto\tau$. The parity-resolved DQPTs and the $\beta$-phase diagram establish two complementary dynamical diagnostics of nonreciprocal quasicrystals, in which nonreciprocity promotes transport at the critical point and suppresses it in the delocalized phase.
Read the full abstract
Applied Physics
Surrogate-accelerated parameterisation of physics-based Li-ion battery models
Physics-based lithium-ion battery models provide access to physically meaningful internal electrochemical states and processes, but cell-specific parameter inference from terminal current-voltage data is computationally expensive and limited by identifiability. We present a surrogate-accelerated inverse framework based on a single-particle model with electrolyte dynamics (SPMe). Its forward map uses our Artiphy surrogate framework for rapid, differentiable evaluation of voltage and selected internal states. After rescaling to remove exact structural redundancies, we infer non-redundant transport, kinetic and capacity parameter groups, including concentration-dependent solid and electrolyte diffusivities. Synthetic voltage data from a Doyle-Fuller-Newman (DFN) model under a WLTP-like current protocol provide a benchmark with known reference parameters and controlled model discrepancy. The inferred SPMe reproduces the benchmark voltage with an error of order 1 mV and recovers electrode capacities well. Positive-electrode diffusivity is recovered accurately over much of the probed stoichiometric range. Local sensitivity and Fisher-information analysis identifies correlated kinetic-Ohmic and electrolyte-transport directions, and shows how localised information and the global diffusivity parameterisation can yield narrow Fisher-curvature envelopes despite weak voltage sensitivity to negative-electrode diffusion over much of the drive cycle. These results represent a step towards rapid physics-based in-silico parameterisation and reduced reliance on destructive cell characterisation.
Read the full abstract
Number Theory
On the syntomic regulator of the Hesse cubic curves and $p$-adic hypergeometric functions
We introduce a new type of $p$-adic hypergeometric function, which satisfies congruence relations similar to Dwork's $p$-adic hypergeometric function. Also, we prove that the syntomic regulators of the Hesse cubic curves are expressed in terms of the special values of our new $p$-adic hypergeometric functions. We also show that there is a transformation formula between our new $p$-adic hypergeometric function and a $p$-adic hypergeometric function of logarithmic type defined by Asakura.
Read the full abstract
Robotics
A Confidence-Aware Multimodal Fusion Framework for Industrial Human-Robot Collaboration
A confidence-aware multimodal fusion framework (CAMF) is proposed to realize reliable human intention prediction for industrial human-robot collaboration. This framework fuses four heterogeneous modalities including object 6D pose, gaze, skeletal motion and IMU-based hand motion. It embeds a confidence-trend-driven dynamic fusion mechanism into BiLSTM to adaptively balance bidirectional temporal features according to real-time modality reliability. A confidence-guided balanced learning strategy combined with a confidence freezing mechanism is further adopted to adjust network gradients dynamically, suppress noise from low-quality modalities and mitigate cross-modal learning bias. A physical platform based on the UR3 collaborative robot is built for experimental validation. Comparative results show that the proposed method reaches an intention recognition accuracy of 91.86% and outperforms existing multimodal fusion approaches in overall performance and stability. It also maintains satisfactory accuracy under low light and partial occlusion interference. In practical assembly tasks, the framework enables proactive and stable human-robot cooperation with strong environmental adaptability.
Read the full abstract
Sound
TimeCues Studio: A Workspace for Music Annotation and Algorithm Prototyping
Multimedia applications require precise music annotation-labeled positions, segments, or loops-placed by hand or algorithmically. Machine-learning algorithms are scalable and effective but need annotated training data, scarce for many tasks. TimeCues Studio is an open-source workspace where algorithm-development teams annotate a music corpus, compare detection algorithms against those annotations, and prototype new ones. Unlike existing tools built for a single track at a time, TimeCues targets teams annotating whole collections, tightly integrated with algorithm development. Annotators place several marker types-each supporting ambiguity-aware labeling-on a grid-locked timeline that visualizes many music features, including separated audio stems. The same timeline drives an algorithm-comparison engine with bundled baselines, a Python sandbox for prototyping new models, and an ambiguity-aware evaluator that honors the structured fields. The same visualization suits solo annotators on music-sync projects. TimeCues is MIT-licensed and deploys via one Docker Compose command.
Read the full abstract
Fluid Dynamics
The effects of surfactant solubility on inertial Marangoni flow: theory and numerics
Surface active agents (surfactants) are common contaminants found on most air-liquid interfaces. Surfactants lower the surface tension, and hence surfactant concentration gradients generate surface tension gradients, which leads to flow. A canonical surfactant-induced flow is the outward spreading due to the localised deposition of surfactants onto an otherwise clean liquid interface. Eshima et al. (Phys. Rev. Lett., accepted, 2026) demonstrated experimentally and theoretically, through a late-time similarity solution, that surfactant solubility enhances surfactant-induced flows of air-liquid-air sheets and that soluble surfactant solutions can be mapped onto equivalent insoluble surfactant solutions. Here we extend the work theoretically and numerically, developing a systematic framework for the derivation and analysis of such solutions. The relevant nondimensional parameters are quantitatively defined, and the physically accessible parameter space is wide, which our theory and numerical simulations are able to span. We link the solutions for finitely soluble surfactants to the solutions for the limiting regimes of insoluble and infinitely soluble surfactants, where the latter link only holds transiently: the solution of the infinitely soluble limit appears as an intermediate solution that persists ever longer as solubility increases, before the ultimate late-time solution, which maps onto insoluble surfactants, is obtained.
Read the full abstract
Robotics
Odometer-Agnostic Drift Correction Using OpenStreetMap Lane Geometry
Despite significant progress in odometry estimation, long-term drift remains a fundamental limitation of incremental pose integration, especially in large-scale or loop-free environments. Existing map-assisted methods can reduce drift, but often depend on dense maps, sensor-specific processing, or complex matching pipelines. We propose a lightweight open-source, odometry-agnostic correction method that aligns short trajectory segments to OpenStreetMap (OSM) lane centerlines. By formulating drift correction as a direct alignment between recent odometry and sparse lane geometry, the method enables efficient online operation without dense priors or expensive preprocessing. Experiments with LiDAR and visual odometry backends demonstrate consistent improvements, with particularly strong gains under severe drift.
Read the full abstract
Artificial Intelligence
From Symbolic Perception to Logical Deduction: A Framework for Guiding Language Models in Geometric Reasoning
Plane geometry remains a significant challenge in AI, requiring the integration of visual perception and mathematical reasoning. While Large Multimodal Models (LMMs) naturally handle visuo-linguistic inputs, they are often computationally intensive and opaque. We demonstrate that a pure Large Language Model (LLM), when equipped with specialized modules, can rival state-of-the-art LMMs on complex geometry problems. Our framework integrates a Geometric Vision Parser, which translates diagrams into symbolic form, with a Symbolic Solver that performs formal deductions, thereby mitigating hallucinations and promoting interpretable reasoning. To enable rigorous evaluation, we curate a benchmark of challenging problems from the 2025 Chinese Zhongkao examinations, ensuring data novelty and testing deeper deductive skills. Experiments demonstrate that our approach achieves performance comparable to Gemini 2.5 Pro while delivering clearer, human-like solutions.