Peer Revenue Past citations are no guarantee of future returns

Papers

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

All of arXiv

Monday 14 September

Computer Vision

Robust Multi-Model Fitting through Learning Neighbor Regions

Multi-model fitting involves fitting multiple models accurately in a noisy environment. It is the basis for computer vision tasks such as scene reconstruction and mixed reality. However, its performance is often limited by insufficient feature utilization, inefficient optimization, model overlap, and the non-differentiable pipelines. To overcome these limitations, we introduce a robust coarse-to-fine framework called Learning Neighbor Regions (LNR). Recognizing that substantial computational resources are wasted on numerous bad minimum sets, we propose the coarse-level module. This module utilizes a neural network to extract and analyze geometric feature of both local point-wise relationships and global contextual information in minimum sets, outputting confidence to pre-select a small number of good minimum sets, thereby enhancing overall efficiency before solving hypotheses. To address model overlap, LNR encodes neighbor region features for each hypothesis in its fine-level module. These region features consist of geometric features of neighboring data points, which can be used by multiple regions simultaneously. This design allows the neural network to individually refine and score each hypothesis. Importantly, LNR is trained to learn directly from data point features rather than from the hypothesis parameters, thus avoiding differentiating the sampling process and the model solvers. Extensive experiments on four classic multi-model fitting tasks demonstrate that LNR achieves state-of-the-art performance. The analysis suggests that LNR can be easily adapted to various robust multi-model fitting tasks.
Read the full abstract

All authors Chang Nie, Guangming Wang, Zhe Liu, Hesheng Wang

General Relativity and Quantum Cosmology

Asymptotic Decoherence of an Unruh–DeWitt Detector in de Sitter Spacetime: Conformal versus Minimal Coupling

We investigate the loss of coherence of a two-level Unruh–DeWitt detector coupled to massless real scalar fields in $(1+3)$-dimensional de Sitter spacetime. Treating the detector–field interaction perturbatively to second order, we derive analytic expressions for the real asymptotic coherence-decay coefficients for conformally and minimally coupled scalar fields. For the conformally coupled field, the coefficient remains finite in the small-gap limit and grows linearly with the detector energy gap in the large-gap regime. For the massless minimally coupled field, the logarithmic sector of the Wightman function produces an additional positive contribution. This provides an explicit analytic extension of the previously known infrared enhancement of detector transition responses to the asymptotic coefficient governing detector coherence. The explicitly average-time-dependent sector does not contribute to the real nonzero-frequency asymptotic coefficient under the adiabatically regulated half-line prescription adopted here, when the regulator is removed at fixed average time and fixed positive detector gap. This asymptotic result does not describe the complete finite-time dynamics, which depends on the switching function and observation interval. Within the stated prescription, the coefficients satisfy $\Gamma_{\rm MMC}>\Gamma_{\rm CC}$ for $\omega>0$, with $\Gamma_{\rm MMC}/\Gamma_{\rm CC} =1+(H/\omega)^2$. The distinction is most pronounced in the small-gap regime, whereas the two coefficients approach one another for large detector gaps.
Read the full abstract

Probability

Occupation BAsed Propagation of Chaos for MacKean Vlasov Forward Backward Systems with Jumps and Monotone Sources

We study robust Mckean-Vlasov Forxard Backward systems with jumps under a nondominated family of probability measures. The backward equation contains a lawdependent nonsmooth monotone source motivated by asymmetric loss aversion. We establish well-posedness and stability of the system, analyze its Yosida regularization, and develop quantitative particle approximations. The main difficulty comes from the discontinuity of the minimal selection across switching interfaces, which prevents the use of direct Lipschitz estimates. To overcome this difficulty, we combine forward propagation of chaos with occupation estimates for the limiting forward process. Importantly, no occupation estimate is required for the particle system. Under a second-moment occupation condition, we obtain an improved joint convergence rate for the particle and Yosida approximations, controlling the forward and backward states, the martingale integrands, and the accumulated source. We also verify the required occupation conditions for several concrete jump models, including affine stable-like dynamics, predictable variable jump amplitudes, and a non-affine class obtained by smooth conjugacy
Read the full abstract

All authors Tcheick Kayembe Tshiswaka, Rostin Mabela Matendo, Simon NTumba Badibanga, Jean-Pierre Bosonga Bofeki Lounga

Quantum Physics

Improving precision scaling via backaction-evading continuous measurement in a driven-dissipative Kerr parametric oscillator

Dissipative phase transitions in the driven-dissipative Kerr parametric oscillator offer a promising route for realizing criticality-enhanced quantum sensing based on continuous measurements. However, achieving such enhancement through realistic measurement schemes remains an outstanding challenge. Here, we extend the backaction-evasion strategy introduced in our earlier work for the Gaussian linear case [arXiv:2511.22248 (2025)] to analyze how the quantum and classical Fisher information scale with the Kerr nonlinearity at dissipative critical points. Our results show that backaction-evading homodyne monitoring achieves enhanced photon-number scaling that surpasses the standard quantum limit, and significantly outperforms alternative protocols such as continuous photon counting. As an additional methodological contribution, we also implement and benchmark time-discrete approximation schemes with improved statistical convergence properties. We use these methods to compute the classical Fisher information for continuous homodyne detection, and demonstrate that they provide efficient access to this quantity near dissipative critical points, thereby extending the reach of existing methods.
Read the full abstract

All authors Cheng Zhang, Xinhui Cui, Jiaying Pan, Xin-Qi Li, Mauro Cirio, Pengfei Liang

Artificial Intelligence

Parameter-Efficient Adaptation of Pretrained Language Models for Time-Series Forecasting

We study the adaptation of pretrained language models to univariate time-series forecasting through a parameter-efficient transfer learning framework, with the goal of understanding which design choices drive effective cross-modal transfer. While language models operate on discrete textual tokens, time series consist of continuous numerical observations with temporal dependencies. To bridge this modality gap, we project fixed-length time-series patches directly into the embedding space of a pretrained GPT-2 backbone, bypassing textual tokenization and treating the Transformer as a generic sequence encoder. Through controlled ablation studies on seven benchmark datasets spanning energy, weather, traffic, and finance, we analyze the effects of (i)~representation strategy (continuous embeddings versus textual serialisation), (ii)~adaptation regime (frozen backbone versus partial or full fine-tuning), (iii)~architectural components such as adapters, pooling strategies, and prediction heads, and (iv)~input context length. Continuous patch-based embeddings consistently outperform textual prompting and randomly initialised backbones. The adapted pipeline attains MASE within the range of specialised forecasting architectures while updating less than 1% of total model parameters. Results further indicate that freezing the pretrained backbone and training lightweight projection and adapter modules provides a favourable accuracy–efficiency trade-off with stable behaviour across varying context lengths.
Read the full abstract

Probability

Optimal Time-Dependent Jump Truncation for Time-Singular Lévy Processes

We study optimal jump truncation for the additive time-singular pure-jump model of the form $X_T=\int_0^T t^{-\sigma} dZ_t$, where $Z$ is a Lévy process and $\sigma\ge 0$. For a fixed expected jump cost, we minimize the residual small-jump variance over measurable time-dependent cutoffs. Under regularity and tail assumptions on the Lévy measure, we prove that the problem admits an optimal cutoff, unique up to a.e. equality, of the form $r^\ast(t)=(c^\ast t^\sigma)\wedge 1$. In the symmetric $\alpha$-stable case, we obtain explicit matched-cost comparisons with the classical fixed cutoff and show that within the admissible non-truncated regime the Dynamic Cutting family is strictly better than the classical fixed cutoff whenever $0<\sigma<1/2$. Finally, for symmetric Lévy measures and cutoffs satisfying the corresponding $L^p$-integrability assumptions, we derive the weak-error bound $W_p(r)\le C_p \mathcal{E}^{p/2}(r)$, $0<p<2$, demonstrating that any cutoff which minimizes the residual variance also minimizes the corresponding upper bound for the weak error.
Read the full abstract

Quantitative Methods

TractSpLearn: Specialized Shared-Manifold Learning for Individualized Detection of Subtle White Matter Alterations in Mild Traumatic Brain Injury

Traumatic brain injury (TBI) often leads to subtle white matter damage that remains undetected on conventional MRI. Diffusion kurtosis imaging (DKI), an extension of diffusion tensor imaging (DTI), provides complementary information on non-Gaussian water diffusion and is sensitive to complex white-matter microstructure. With the advent of ultra-high-field MRI, the spatial resolution and signal-to-noise ratios (SNR) have been significantly enhanced, enabling more precise visualization of subtle abnormalities. Building on these advances, we developed TractSpLearn, an individualized tract-based learning framework that jointly considers within-group variability and between-group differences. Unlike the original TractLearn framework, which learns a normative manifold exclusively from healthy controls, TractSpLearn incorporates both healthy controls and patients to learn a shared manifold with a healthy-anchored representation and an additional patient-related component. To assess the performance of the proposed method, we compared TractSpLearn with the original TractLearn in three cohorts: (i) healthy controls (HC), (ii) athletes with persistent post-concussive syndromes (PPCS), and (iii) athletes with repeated head injuries (RHI), with abnormalities particularly evident in axial kurtosis (AK) and mean diffusivity (MD). In RHI, TractSpLearn highlighted recurrent abnormalities across patients. In the PPCS cohort, the overall group-level differences were more modest, potentially reflecting both limited statistical power due to the small sample size and partial normalization of white-matter alterations during recovery. Still TractSpLearn identified abnormality evidence in more patients and across more affected tracts than TractLearn.
Read the full abstract

All authors Jiqing Huang, Ali Al-Husseini, Yi Chen, Anna Gard, Laurent Lamalle, Mohamed Ali Bahri, Markus Nilsson, Niklas Marklund, Christophe Phillips, Evgenios N. Kornaropoulos

Systems and Control

Communication Architecture Co-Design for Distributed Control via System Level Synthesis

As networked dynamical systems grow in scale, distributed controller architecture design has become a critical issue. To address this, prior studies have incorporated the regularization for design (RFD) framework into system-level synthesis (SLS), focusing primarily on actuator and sensor placement. However, communication architectures also fundamentally impact the performance of distributed control, and scalable explicit co-design of the underlying communication topology within SLS remains challenging. This paper proposes a scalable convex framework for the co-design of communication architectures and distributed controllers within the SLS paradigm. We explicitly characterize a communication topology-induced feasible subspace of system responses, parameterized by the communication adjacency matrix and separate from the physical plant topology. By deriving a communication link norm from this characterized subspace, we formulate a convex SLS problem for communication architecture co-design. A proximal ADMM scheme is employed to appropriately decompose the optimization problem in a manner that preserves scalability to large networks. Numerical experiments validate the efficacy of the proposed framework in optimizing the trade-off between closed-loop performance and communication architecture complexity, while also confirming its scalability.
Read the full abstract

Materials Science

Optical properties of single-crystal As2Se3 from first principles

Crystalline As2Se3 is a promising layered chalcogenide for mid-infrared photonics and optoelectronic applications, yet its fundamental electronic and optical properties remain debated. This paper combines density functional theory with G0W0 quasiparticle corrections and Bethe-Salpeter equation (BSE) calculations to deliver a robust description of its band structure and optical response. The G0W0 results reveal an indirect band gap of 2.31 eV with nearly degenerate direct transitions at Gamma, while the BSE spectra exhibit strong polarization-dependent anisotropy and pronounced excitonic effects. Several bound excitons are identified near the absorption onset, clarifying the role of electron-hole interactions and resolving previous controversies regarding the nature of the optical gap and the contribution of lone-pair states. This work provides a unified and quantitatively accurate picture of As2Se3, reinforcing its potential for next-generation infrared and anisotropic photonic devices.
Read the full abstract

Neural and Evolutionary Computing

A Memristive Synapse for Online STDP Learning and Inference in SNNs

This work presents a fully analog memristive synaptic circuit for online spike-timing-dependent plasticity (STDP) learning in spiking neural networks (SNNs). The proposed synapse integrates a local STDP circuit generating gradual timing-dependent conductance updates directly from pre- and post-synaptic spikes. Learning occurs during normal network operation without requiring external digital control or explicit STDP waveform synthesis. Post-layout simulations of the memristive synapse implemented in a 130 nm CMOS technology show spike-timing-dependent conductance adaptation during SNN operation. A 2x2 SNN simulation further illustrates online neuron specialization through unsupervised learning.
Read the full abstract

All authors Elia Mateu-Barriendos, Álvaro Gómez-Pau, Daniel Arumí, Rosa Rodríguez-Montañés, Salvador Manich

Computation and Language

Dynamic Semantic Compression for Efficient Latent-Space Inference in Large Language Models

Large Language Models (LLMs) primarily perform inference at the token level, resulting in substantial memory overhead and compromised computational efficiency. In this paper, we propose a Dynamic Semantic Extraction and Inference (DSEI) framework, which achieves segment-level inference within the latent space through a two-stage training strategy. First, we construct a Dynamic Semantic Autoencoder (DSAE) via self-supervised learning. DSAE dynamically extracts segment-level semantics and compresses them into compact latent representations via adaptive semantic weighting and gated fusion. Subsequently, we integrate the DSAE into the LLM architecture and train the model to infer over dense latent space. DSEI substantially reduces both input and generation sequences and significantly enhances inference efficiency. Extensive experiments conducted on the Wanjuan dataset demonstrate that DSEI reduces perplexity by 48% compared to static sentence-level latent inference baseline. Furthermore, compared to standard LLMs using token-level inference, DSEI accelerates inference speed by 2.5$\times$ and reduces memory overhead by 90%.
Read the full abstract

All authors Peipei Li, Dongsen Zhang, Yuchen Liu, Wenjun Xu

Optimization and Control

Directionally Fine Families of Cones: Construction Principles, Approximation, and Separation

We introduce the notion of a directionally fine family of cones, a local geometric property expressing the possibility of selecting arbitrarily narrow cones around every direction of a normed space. We show that this local property provides a common mechanism for finite separation and approximation. In particular, directionally compact sets can be separated from closed cones by finitely many members of the family, while locally compact cones admit finite outer approximations with quantitative Hausdorff control on their sets of directions and on bounded sections. These finite constructions lead naturally to max-type positively homogeneous separators and do not require convexity of the cones to be separated. We then study two complementary realizations of the framework. For the classical family of Bishop–Phelps cones, directional fineness is characterized by the requirement that every point of the unit sphere be a denting point of the unit ball, equivalently, by property~$(G)$; in Banach spaces this amounts to rotundity together with the Kadec property. On the other hand, we introduce transversal coercivity as a general axial construction principle. It yields directionally fine families in arbitrary normed spaces, without geometric assumptions on the norm, and includes uniform axial deviation and norm-normalized axial cones as canonical models. Finally, the explicit structure of the former yields penalization and recovery results for optimization over locally compact cones, together with quantitative approximation and convergence estimates.
Read the full abstract

Complex Variables

Optimal Linear Dependence on Boundary Type for Local Gromov Hyperbolicity of the Kobayashi Metric

The Gromov hyperbolicity constant of a metric space \((X,d)\) is the infimum of all \(\delta\ge0\) such that \((X,d)\) is \(\delta\)-hyperbolic. For a Kobayashi hyperbolic domain \(\Omega\subset\C^n\) and a boundary point \(p\in\partial\Omega\), let \(\delta_{\mathrm{loc}}(\Omega,p)\) denote the local Gromov hyperbolicity constant obtained by restricting the points to arbitrarily small Euclidean neighborhoods of \(p\), while distances are still measured by the ambient Kobayashi distance \(K_\Omega\). For each even integer \(M\ge2\), let \(\mathfrak H(M)\) be the supremum of these constants over all complex dimensions \(n\ge2\) and all domains whose boundary is smooth and convex near the distinguished point and has D'Angelo type at most \(M\) at that point. We prove \[ \frac{\log 2}{2} M \le \mathfrak H(M) < 36M. \] Thus the optimal universal dependence of the local Gromov hyperbolicity constant on boundary type is linear. We also show that no analogous bound holds for the global Gromov hyperbolicity constant, even among smooth bounded strongly convex domains.
Read the full abstract

All authors Cheng Lou, Jianyong Qiao, Hongyu Wang, yumin Zhong

Computer Vision

Query-Conditioned Spherical Centroid Aggregation for Multimodal Retrieval

Multimodal retrieval integrates video, audio, subtitles, and text; however, recent geometric aggregators, such as Gramian volumes, hyperbolic volumes, and spectral objectives, treat all modalities symmetrically. Under a unified evaluation protocol, their joint scores frequently lag behind the strongest single-modality pathway by 1.9 to 27.6 R@1. Controlled analyses attribute this outcome to uniform modality influence. This work introduces Spherical Centroid Aggregation with Learned Adaptive Relevance (SCALAR), a query-conditioned aggregator that assigns relevance-based weights to each available modality before computing a spherical centroid. SCALAR accommodates arbitrary modality subsets and is trained on masked, reduced-arity views using rank-8 LoRA adapters. Across five benchmarks, SCALAR achieves positive aggregation gain on four, reaching +4.0 R@1, while none of the evaluated prior aggregators is positive on more than one. A uniform-weight ablation reproduces the degradation observed with symmetric aggregation. With only 4.8 million trainable parameters, SCALAR attains the highest text-to-video R@1 on three and performs within seed variation of the best result on a fourth. Under test-time modality dropout, SCALAR's representation-stage score surpasses the released GRAM checkpoint at every evaluated masking rate and benchmark by 3.2 to 10.9 R@1. Finally, as modalities are removed, rerankers trained exclusively on complete modality sets increasingly converge toward their video-only pathways, diminishing these representation-level gains and underscoring a limitation of standard two-stage retrieval pipelines.
Read the full abstract

Computer Vision

Concept-Grounded Reasoning with Prompt-Driven Localization for Interpretable Structured Report Generation

Medical imaging modalities such as ultrasound and X-ray are widely used in clinical practice, where diagnosis follows a structured, evidence-driven workflow aligned with standardized criteria. While multimodal large language models (MLLMs) show promise for automated medical report generation, most existing systems rely on end-to-end multimodal fusion without modeling clinically defined intermediate attributes, leading to limited grounding and interpretability. To address this issue, we propose CORAL (COncept-grounded ReAsoning with Localization), a multimodal framework that integrates spatial grounding and concept-level supervision into a unified reasoning process. CORAL employs a prompt-driven medical segmentation model to localize lesions and predicts multi-class clinical attributes through a Concept Bottleneck module. The resulting textual concept tokens are combined with mask-modulated visual features within an MLLM to enable structured report generation and diagnostic prediction. Experiments on BUS-CoT and IU X-ray datasets demonstrate consistent improvements in diagnostic accuracy, concept consistency, and report quality over strong general-purpose and medical MLLMs, indicating that concept-grounded reasoning better aligns generation with clinical decision processes.
Read the full abstract

All authors Xinyue Xu, Hongbin Lin, Juangui Xu, Hualiang Wang, Lehan Wang, Lijie Hu, Adrian Weller, Xiaomeng Li

Algebraic Geometry

An algebraic comparison of $\widehat{\mathrm K}$-stability and $\mathrm K^\beta$-stability

We establish a uniform quantitative comparison between the non-Archimedean Mabuchi functional and the Darvas–Zhang quantised Mabuchi functional. For a smooth polarised variety with finite automorphism group, we provide an algebraic proof that uniform $\widehat{\mathrm K}$-polystability is equivalent to uniform $\mathrm K^\beta$-stability for every sufficiently large rational $\beta>1$, without passing through the existence of a cscK metric. The main input is a square-root estimate for directional derivatives of non-Archimedean energies, proved by intersection theory. We also study a reduced $\mathrm{K}^\beta$-stability condition when the identity component of the polarised automorphism group is reductive and the Futaki character vanishes. We conclude by characterising K-semistability in terms of asymptotic $\mathrm K^\beta$-semistability and prove the quantitative convergence of the corresponding stability thresholds.
Read the full abstract

Functional Analysis

The BSE Property for Weighted Banach Algebras of Dirichlet Series

Let $\omega: \mathbb{N} \to [1, \infty)$ be a weight on the multiplicative semigroup of natural numbers. We investigate the Bochner-Schoenberg-Eberlein (BSE) property for the weighted Banach algebra of Dirichlet series $A_\omega = \ell^1(\mathbb{N}, \omega)$ under Dirichlet convolution. By employing Bohr's correspondence, we establish a weak-$\ast$ approximation framework for $\omega$-bounded semicharacters using scaled truncations that lie within the predual space $c_0(\mathbb{N}, 1/\omega)$. Utilizing the canonical $L$-projection decomposition of the bidual space $A_\omega^{\ast\ast} = A_\omega \oplus c_0(\mathbb{N}, 1/\omega)^\perp$ alongside the weak-$\ast$ compactness of the unit ball via the Banach-Alaoglu theorem, we prove that $A_\omega$ is a BSE-algebra for any weight. Furthermore, we show that the intrinsic BSE-norm on the Gel'fand transforms is isometric to the natural weighted algebra norm $\|\cdot\|_\omega$.
Read the full abstract

Solar and Stellar Astrophysics

SPAMMS: 3D spectroscopic modeling of stellar surfaces. II. Implementation of Kurucz and TLUSTY model atmospheres

Context. Accurate stellar spectra are essential to derive stellar properties. Traditional model atmospheres often oversimplify phenomena that break spherical symmetry, such as rotational deformation or multiplicity. The Spectroscopic PAtch Model for Massive Stars (SPAMMS) accounts for these effects, but its applicability has been limited by the spectral types covered by its model atmosphere grids. Aims. We aim to extend the parameter space of the model atmosphere grids available to SPAMMS, enabling spectral synthesis across a broader range of stellar types. Methods. We computed specific intensities, $I\left(\lambda,\mu\right)$, for $101$ emergent angles using PRISMAS (Pipeline of Radiative Intensity Synthesis for Meshed Atmospheric Surfaces), and pre-computed LTE and non-LTE atmospheres from two ATLAS9-Kurucz grids and the TLUSTY-based OSTAR2002 and BSTAR2006 models. Results. The intensity grids cover effective temperatures from $3500$ to $55000\,\mathrm{K}$ and surface gravities from $0.0$ to $5.0\,\mathrm{dex}$, spanning O- to K-type stars. They include metallicities from $0$ to $30\,\mathrm{Z_\odot}$ and microturbulent velocities of $1$, $3$, $5$, and $10\,\mathrm{km\,s^{-1}}$. The spectral range extends from $3000$ to $9000\,\mathring{\mathrm{A}}$ with $\Delta\lambda=0.01\,\mathring{\mathrm{A}}$. As a proof of concept, we modelled with SPAMMS a rapidly rotating B-type star and an eclipsing Algol-type binary. Conclusions. The new LTE-Kurucz and NLTE-TLUSTY grids substantially expand the parameter space accessible to SPAMMS. The code can now generate synthetic spectra for a broader range of stellar types and geometries, including rapidly rotating and multiple systems – providing a more comprehensive framework for modelling non-spherical stellar surfaces.
Read the full abstract

All authors D. Galán-Diéguez, M. Abdul-Masih, C. Allende Prieto, S.R. Berlanas, A. Herrero, H. Sana

Combinatorics

On Erdős Problem 767: Cycles with Chords

For integers $k\ge 1$ and $n\ge k+2$, let $g_k(n)$ be the maximum number of edges in an $n$-vertex graph containing no cycle with a vertex incident with at least $k$ chords. Erdős conjectured that $g_k(n)=(k+1)(n-k-1)$ for $n\ge 2k+2$. Lewin found a counterexample. Bollobás later conjectured that there exists a function $n(k)$ such that $g_k(n)=(k+1)(n-k-1)$ for all $n\ge n(k)$. Jiang confirmed this by proving the formula for all $n\ge3k+3$ when $k\ge1$. In this paper, we determine $g_k(n)$ completely. For all $k\ge1$ and $n\ge k+2$, we prove $g_k(n)=\big\{\lfloor\frac{(k+1)n}{2}\rfloor,\max\{a(n-a)+\lfloor\frac{a(k+1-a)}{2}\rfloor : a\in\mathbb Z,\; \lfloor {(k+1)}/{2}\rfloor+1\le a\le k+1\}\big\}$. For $k\ge2$, we prove $g_k(n)=(k+1)(n-k-1)$ when $n\ge \lceil(5k+1)/2\rceil$, and this threshold is sharp. Our proof builds on the method developed by Ma and the second author in [Ma and Ning, 2020].
Read the full abstract

Number Theory

Optimising Selberg's method for critical zeros

We revisit Zhuravlev's 1974 quantitative form of Selberg's sign-change method for the zeros of the Riemann zeta function on the critical line. Zhuravlev's work, originally in Russian and little known in the West, established an explicit positive proportion of such zeros. Using modern techniques we optimise the arithmetic mean value at the heart of the method and obtain the strongest result yet known from Selberg's method, proving that more than $6.78\%$ of the non-trivial zeros of the Riemann zeta function lie on the critical line.
Read the full abstract

Number Theory

On the arithmetic of Tate-Shafarevich groups via Kolyvagin's conjecture

We investigate how Heegner points detect the finiteness of the $p$-primary part of Tate-Shafarevich groups of semi-stable elliptic curves over the rationals of arbitrary rank with any good reduction prime $p \geq 5$ with irreducible mod $p$ representation and partially recover Çiperiani–Wiles' theorem on solvable points on genus one curves. As a key ingredient of the proof of these results, we also give a concise proof of Kolyvagin's conjecture for semi-stable elliptic curves with supersingular reduction with a relevant choice of an imaginary quadratic field. Our approach is independent of the cyclotomic Iwasawa main conjecture for elliptic curves with supersingular reduction.
Read the full abstract

Atomic Physics

Diagonal Born–Oppenheimer correction in strong magnetic fields: finite-difference approach for light diatomic molecules

The influence of the diagonal Born–Oppenheimer correction (DBOC) on the electronic structure of light diatomic molecules subjected to strong magnetic fields is investigated. H$_2$, HeH$^+$, and LiH are considered, using both the Hartree–Fock (HF) approximation and configuration interaction with single and double excitations (CISD). The analysis covers magnetic field strengths up to $B_\parallel = 0.2$ a.u. ($4.7\times10^{4}$~T), relevant to astrophysical conditions near magnetic white dwarfs. Although the correction noticeably shifts total electronic energies, it varies smoothly with internuclear distance and depends only weakly on the field, so that it enters the vibrational transition frequencies as a nearly field-independent offset, shifting the fundamental interval by about $1$~cm$^{-1}$ and the higher levels by up to $15$ cm$^{-1}$. These shifts exceed the $0.1$–$2.0$ cm$^{-1}$ accuracy currently attainable for molecular lines in the atmospheres of magnetic white dwarfs. Comparison of HF and CISD results shows that electron correlation has a non-negligible effect on the DBOC, most pronounced for the strongly ionic LiH bond.
Read the full abstract

Emerging Technologies

System-Technology Co-Evaluation of A7 CFET and A10 NSFET Technologies from Cell Parasitics to Chip Reliability

Complementary FETs (CFETs) extend nanosheet FET (NSFET) scaling by vertically stacking n- and p-type gate-all-around (GAA) devices, thereby shrinking standard-cell area. The performance gain, however, cannot be assessed from device metrics alone, as CFET layouts also introduce larger cell-level parasitic resistance and capacitance (RC). In this work, we present a physics-based thermal- and aging-aware system-technology co-evaluation (STCO) flow to assess parasitic RCs in A7 CFET and A10 NSFET technology nodes. Our flow links calibrated device models, optimized standard-cell generation, automated GDS-to-TCAD conversion enabling accurate 3D parasitic RC extraction, full RTL-to-GDS implementation for an AI accelerator, multiphysics thermal analysis, and physics-based bias temperature instability (BTI) aging evaluation. Using the same device model for both technologies, we can isolate the impact of parasitic RCs and design at different levels of the design flow. The results of the AI accelerator design demonstrate that the A7 CFET reduces the chip area by 24.7% and the total wire length by 12%, improving the area efficiency TOPS/mm^2 by 74% relative to the baseline of the A10 NSFET. Under iso-frequency operation, results reveal that CFET voltage scaling reduces power by 68% and lowers power density from 148 W/cm^2 to 55 W/cm^2, which reduces the chip's temperature from 125 degrees C down to merely 62 degrees C. The resulting reduction in stress temperature suppresses 10-year BTI-induced degradation by 39%, reducing the required aging timing guardband by 53%.
Read the full abstract

All authors Mahdi Benkhelifa, Leon Mayr, Hadi Nour Eddine, Andrea Padovani, Luca Larcher, Hussam Amrouch

General Relativity and Quantum Cosmology

Cosmological dynamics and structure formation in Tsallis-Cirto entropy-inspired modified gravity

We explore the cosmological dynamics and formation of structures within the Tsallis-Cirto modified gravity framework with a model parameter $\beta$ that is inspired by nonextensive statistics. By applying the gravity-thermodynamics conjecture, we modify the Friedmann equations and the evolution of the Hubble parameter. The impact of the Tsallis-Cirto entropy on different cosmographic parameters and the development of the linear matter overdensities have been analyzed by formulating perturbed field equations using the spherical collapse approach in a flat Friedmann-Lemaître-Robertson-Walker background. We present a novel and established diagnostic approach to distinguish among distinct cosmological models compared to flat and non-flat $\Lambda$CDM scenarios, discovering that the Tsallis-Cirto entropy-inspired modified cosmology ($\beta\ne 1$) successfully withstands all tests, contradicting both the flat and non-flat $\Lambda$CDM models. This model also meets the conditions for the Universe to reach thermodynamic equilibrium in the far future. Additionally, we investigate the halo mass function and cluster number counts within this modified gravity framework. All findings are compared to the fiducial $\Lambda$CDM profile, indicating that the added entropic correction affects the history of expansion, the growth rate of structures, and the dark matter halo abundance. We find that the more massive structures are less abundant and develop at later times, which aligns with the hierarchical model of formation of large-scale structures.
Read the full abstract

High Energy Physics (Phenomenology)

Vertexing displaced diphoton decays with recoil photons at Belle II

We propose a recoil-assisted strategy for vertexing displaced diphoton decays in $e^+e^-\to\gamma X$, $X\to\gamma\gamma$, using only one converted daughter photon. The initial state and recoil-photon momentum define the LLP flight line, whose closest approach to the converted-photon trajectory locates the decay vertex. This removes the need for a second conversion, making the conversion-related efficiency scale linearly rather than quadratically with the photon-conversion probability. For photophilic axionlike particles at Belle II, the existing $408$ fb$^{-1}$ data set could probe previously unconstrained parameter space near $m_a \simeq 100$ MeV and $g_{a\gamma\gamma} \simeq 2 \times 10^{-4}$ GeV$^{-1}$, improving leading bounds by almost an order of magnitude and testing the mass region below the $\pi^0$ mass, which the existing Belle II three-photon searches cannot access. With $50$ ab$^{-1}$, the reach extends to $m_a \simeq 310$ MeV and $g_{a\gamma\gamma}\simeq2\times10^{-5}$ GeV$^{-1}$.
Read the full abstract

Disordered Systems and Neural Networks

Statistical levels and spatial modes of Fock-space heterogeneity in many-body localization crossovers

Near many-body localization crossovers, local memory fluctuates strongly among eigenstates and ensemble realizations, but the observed heterogeneity combines contributions from different statistical levels. We develop a statistical framework based on configuration-space distance distributions that uses a variance decomposition to separate fluctuations within eigenstates, between eigenstates of one sample, and between samples. Applying this framework to random-cosine and quasiperiodic-cosine Ising ensembles with the same one-site field marginal, we find that, at the exact-diagonalization sizes studied, the clearest difference occurs among disorder realizations or quasiperiodic phase samples, whereas within-eigenstate and within-sample eigenstate-to- eigenstate contributions remain broadly comparable. Spatial covariances show that random outer fluctuations have a much stronger uniform component, whereas quasiperiodic phase fluctuations are organized more strongly at finite wave number and partly cancel in the spatial average controlling the distance center. We find that, near the crossover in the random ensemble, the ensemble-averaged distance center is particularly sensitive to changes in the nominal field strength. Combined with sample-to-sample differences in the realized field amplitude, this mean response accounts for much of the sample-to-sample variation in the distance center. Analysis of half-chain entanglement further shows that its sample-to-sample fluctuations likewise reflect the combined effects of amplitude variations and its own mean response. An application to a fixed-magnetization spin chain demonstrates the framework in a constrained configuration space. Resolving both statistical level and spatial mode therefore provides a more complete picture of sample- dependent many-body memory and its configuration-space probability geometry.
Read the full abstract

Robotics

Planning in the Backbone: DiffAdapterVLA for Native Continuous Trajectory Generation with Driving VLMs

Pretrained driving vision-language models (VLMs) integrate visual, route, language, and driving context into rich driving priors, yet their representation objectives remain separated from continuous driving planning. Existing methods typically begin trajectory generation only after the VLM has formed a final condition, leaving depth-wise condition computation outside the stepwise formation of trajectory state. We introduce DiffAdapterVLA, which realizes Planning in the Backbone: it injects explicit trajectory tokens into selected VLM late layers, bringing trajectory state into backbone forward computation, where it co-evolves with driving conditions at different depths. Lightweight layer-wise DiffAdapters organize this computation into recursive trajectory refinement, while asymmetric joint attention preserves directed guidance from the condition stream to trajectory planning. By placing planning within existing backbone computation rather than relying on an independent trajectory planner, DiffAdapterVLA adapts only lightweight trajectory modules to turn existing driving priors into efficient continuous planning capability. NAVSIM results show that it achieves high-quality closed-loop planning with low end-to-end latency using few trainable parameters, and demonstrate that jointly evolving trajectory state and depth-wise driving conditions in VLM late-layer computation effectively realizes continuous trajectory planning.
Read the full abstract

All authors Changxin Lu, Xiaoliang Meng, Yu Wu, Rui Huang, Honglin Li, Tao Chen, Kaixuan Zhou, Yadong Shao

High Energy Physics (Phenomenology)

Confronting the Higgsino Interpretation of the LZ Event with Astrophysical Uncertainties and the Solar Capture Constraints

We investigate whether astrophysical uncertainties in the local dark matter distribution can alleviate tensions between the higgsino interpretation of the LZ event and constraints from solar-capture neutrinos. We calculate the expected LZ signal over a wide range of Standard Halo Model parameters and using $N$-body hydrodynamical simulations of Milky Way-like galaxies that include the influence of the Large Magellanic Cloud. For the Standard Halo Model, the requirement $\delta \gtrsim 566~\mathrm{keV}$ implies $v_0 \gtrsim 240~\mathrm{km\,s^{-1}}$, independent of the Galactic escape speed. In the simulated halo, we find that $\delta = 514~\mathrm{keV}$ is the largest mass splitting capable of producing one event in LZ's $200~\mathrm{keV}-300~\mathrm{keV}$ recoil-energy range. We conclude that astrophysical uncertainties are insufficient to reconcile the higgsino interpretation of the LZ event with solar-capture neutrino constraints.
Read the full abstract

Computer Vision

Hypergraph-Regularized Gramian Volumes for Multimodal Retrieval

Volume-based multimodal retrieval jointly scores a text query with a candidate's video, audio, and subtitle embeddings. While this approach captures higher-order within-candidate alignment, the score remains candidate-local, and semantically related training samples primarily serve as contrastive negatives. This work introduces Hypergraph-Regularized Gramian Volumes (HyVol), a training-time module that incorporates these semantic relations prior to evaluating the original volume loss. Document hyperedges connect the observed modalities of each candidate, whereas semantic hyperedges link candidates whose detached captions are mutual top-k neighbors. A shallow gated hyper-graph network applies residual corrections to the modality embeddings. Presence masks exclude unavailable streams from message passing, and identity padding preserves the determinant of the observed Gram submatrix without feature imputation. As refinement operates on embeddings rather than scores, the same construction applies to both Gram and HyperGram. We remove the hypergraph after training, leaving the backbone-only architecture, original scoring function, and retrieval cost unchanged. We train both backbones on a 150K-clip subset of VAST-27M and evaluate zero-shot performance on six benchmarks. Under the paired protocol, HyVol improves R@1 across all five retrieval benchmarks, with video-to-text gains reaching +8.3 on MSR-VTT and +7.6 on VATEX. Under missing-modality masking, the V2T margin remains positive in all experimental settings, although the T2V margin becomes slightly negative in four.
Read the full abstract

Information Retrieval

Clean Scores, Buried Evidence, and Confident Wrong: A Receipt-Based Audit of Frontier Agentic QA

Frontier models score well on shallow document/chart reading tasks. In a controlled data-room audit, moving evidence into buried conditions reduced accuracy, increased forced declarations, increased tool calls, and increased cost per correct answer. Confidence and benchmark calibration did not fully capture wrong answers; a documented production incident shows fabricated structural claims can be mixed with accurate numeric tables. Agentic evaluations need claim-level receipts (statement-level provenance, not answer-level scores), condition-aware scoring, and human-adversarial verification - an auditing discipline, not a leaderboard. The setting we measure is financial due diligence; the setting we are building toward next is defense staff work, where the same buried-evidence shape appears. In both, the model is not a party to the consequences; the person who signs is. In plain terms: in the documented cases we examine, agents can pair accurate numbers with confident fabricated explanations, and the burden of proof must therefore move from the model to the evidence trail.
Read the full abstract