Populations and Evolution
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Friday 11 September
Statistical models of ecological and evolutionary dynamics often include latent variables that are either continuous (e.g., average body size) or discrete (e.g., numerical abundance). Mixed-type hierarchical models containing both are typically fitted using Markov chain Monte Carlo (MCMC), which can be prohibitively slow for large models. Here, we introduce an alternative in the R package RTMB that automates the sequential reduction of small groups of related discrete variables, allowing them to be efficiently marginalized. Sequential reduction is combined with automatic differentiation and the Laplace approximation to estimate parameters and predict both continuous and discrete variables. We demonstrate speed and flexibility using demographic examples (occupancy, dynamic occupancy, N-mixture, and open dynamic N-mixture models), benchmarking RTMB against JAGS and unmarked. We then develop two novel applications. The first is a multi-site open N-mixture model with a spatial latent variable governing site-specific initial abundance and recruitment, which shows that continuous Gaussian Markov random fields can be estimated jointly with discrete abundance dynamics in under a minute. The second is phylogenetic trait imputation for a published data set of female Liolaemus lizards, where we jointly impute a binary trait (viviparity), estimate its state-switching rates, and estimate its effect on a continuous trait (body size) during ancestral state reconstruction. This indicates that phylogenetic comparative methods can estimate linkages among discrete and continuous traits. We envision that intuitive and efficient specification of mixed-type models will allow more expressive representation of ecological and evolutionary dynamics.
Computer Vision
Global-Local Contextual Progressive Expansion Network for Martian Landslide Segmentation in Multimodal Remote Sensing Imagery
Automated landslide segmentation on Mars is one of the important tasks for understanding its surface processes, and all will aid in future space exploration. However, it remains a relatively underexplored open challenge because landslide morphology is highly variable, foreground regions are often sparse or irregular, and orbital observations combine heterogeneous spectral and topographic cues. In this context, this work investigates the capability of deep learning to address Martian landslide segmentation through an extensive assessment of modern neural segmentation models. To the best of our knowledge, this is the first study to provide such a comprehensive exploration in this domain. We further propose TransCPLES, a U-shaped network that couples Contextual Progressive Layer Expansion feature extraction with Transformer-based contextual reasoning, enabling the model to capture local geomorphic patterns and broader spatial dependencies for more reliable landslide delineation. Experiments on MMLSv2, a seven-band multimodal Martian landslide dataset, show that TransCPLES achieves the best overall performance when evaluated on geographically distinct samples, with consistent delineation across different landslide extents, stable foreground discrimination, and a favorable balance between accuracy and computational cost compared with several state-of-the-art convolutional, attention-based, and Transformer-based segmentation models. With this work, we hope to provide a useful reference and encourage further research and development in deep learning for planetary remote sensing. Code will be available after publication.
Read the full abstract
Combinatorics
Spectral extremes under exact cycle conditioning
Let $P_n$ be the matrix of a random permutation of $n$ symbols and let $M_n=\log\max_{|z|=1}|\det(I-zP_n)|$. Cook and Zeitouni proved that $M_n/\log n$ converges in probability to a constant $x_0$ for a uniform permutation. We show that the $\sqrt{\log n}$ fluctuations of $M_n$ are carried entirely by the number of cycles $K_n$. Write $\lambda(s)=\log\{\Gamma(1+s)/\Gamma(1+s/2)^2\}$, let $s_\kappa$ minimize $(1+\kappa\lambda(s))/s$ on $(0,\infty)$, and put $v(\kappa)=\kappa\lambda'(s_\kappa)$ and $a_\theta=\lambda(s_\theta)/s_\theta$. Under the Ewens measure with any fixed parameter $\theta>0$ we prove $M_n=v(\theta)\log n+a_\theta(K_n-\theta\log n)+O_P(\log\log n)$, so that the standardized pair $(K_n,M_n)$ converges jointly to $(G,G)$ with $G$ standard normal: the maximum and the cycle count are asymptotically perfectly aligned. This is deduced from a statement about the exact conditional law, which does not depend on $\theta$: for every compact $[\kappa_-,\kappa_+]\subset(0,\infty)$ there is a finite $C$ such that $P(|M_n-v(k/\log n)\log n|>C\log\log n \mid K_n=k)$ tends to $0$ uniformly over integers $k$ with $\kappa_-\log n\le k\le\kappa_+\log n$, that is, over exact and possibly atypical cycle counts. The proof keeps the size and the cycle count simultaneously in a two-variable coefficient extraction. Cycles longer than $n/(\log n)^4$ are reserved as an analytic factor whose coefficients are flat under every size shift produced by the shorter cycles; positivity then converts a scalar coefficient asymptotic into a relative comparison of the entire path-constrained measure, with an error that does not degrade with the number of constraints or with the rarity of the event. The constrained lower bound comes from pointwise saddle estimates for killed convolutions along a dyadic chain of endpoint boxes.
Read the full abstract
Soft Condensed Matter
Interparticle radiation forces beyond spherical particles in a bulk acoustofluidic device
Understanding interparticle radiation forces between spherical and nonspherical particles is crucial for the trapping of micro- and nano-objects in acoustophoresis. Most theoretical and numerical models treat spherical particles as identical to estimate interparticle forces, owing to their symmetry. Here, we elucidated interparticle radiation forces between three particles with different shapes and material contrast using experimentally validated numerical models. We performed a frequency sweep of the full acoustic device, accounting for viscous effects in a two-dimensional model, to predict the system's resonance. The pressure field obtained at the resonant frequency was then fitted to a 1D sinusoidal wave and used in a three-dimensional model based on the perturbation technique and tensor integral method to calculate the interparticle forces on the particles due to scattering and re-scattering of the acoustic field. Our results revealed a shape-dependent reversal of interparticle forces, even for positive-contrast materials. We found that for spherical particles, the interparticle forces are independent of material properties when the particles are positioned on the nodal plane. However, material contrast and particle asymmetry significantly affect interparticle forces when particles are placed outside the nodal plane. Finally, we studied the effects of the orientation of the asymmetric particle placed outside the nodal plane on the interparticle radiation forces. Interestingly, for the orientation angle of 0 degrees, the interparticle force is maximum for the asymmetrical particle and then gradually decreases to zero as we increase the orientation angle to 90 degrees. The detailed analysis presented in the paper will facilitate a better understanding of shape-based acoustic manipulation of microparticles.
Read the full abstract
Information Theory
When Do Pilots Matter for OFDM Sensing? Pilot-Data Resource Design for ISAC
Practical OFDM-based integrated sensing and communication (ISAC) signals contain deterministic pilots and random data payloads, yet how these two types of resources jointly affect matched-filter sensing performance remains insufficiently understood. This paper establishes an analytical and optimization framework for pilot-data (P-D) OFDM sensing. We first derive a closed-form mean-square periodic autocorrelation function (MS P-ACF), which characterizes the respective contributions of the average subcarrier power spectrum and data-symbol randomness to sensing sidelobes. Based on this result, we reveal that pilot placement has no impact on normalized sidelobes under uniform power allocation or on the normalized global expected integrated sidelobe level (EISL), whereas it becomes an effective design degree of freedom for reshaping sidelobes under nonuniform power allocation. We further derive the globally optimal two-level pilot-data power allocation for global EISL minimization and characterize pilot placement for peak sidelobe suppression. For region-of-interest (ROI) sensing, we jointly optimize pilot placement and subcarrier power allocation to minimize the normalized ROI-EISL. A strictly uniform pilot placement is proved globally optimal under specified conditions, while a bisection-and-branch-and-bound algorithm is developed for general resource configurations. Numerical results validate the theoretical analysis and demonstrate that the proposed design reduces the ROI sidelobe level by up to 2.19 dB compared with optimized uniform-pilot OFDM, while improving ranging accuracy at an SNR of -10 dB by 84.3$\%$. These results reveal when and how pilot placement can be exploited to regulate the sensing behavior of random OFDM communication signals, providing practical resource-design principles for OFDM-based ISAC.
Read the full abstract
Computer Vision
Pedestrian Crossing Intent Classification From Event-Based Vision Using Convolutional Spiking Neural Networks With Temporal Augmentation
Anticipating whether a pedestrian will cross the road is safety-critical for autonomous vehicles, requiring real-time inference under challenging conditions including motion blur, high dynamic range, and class imbalance. Conventional frame-based deep networks process redundant RGB data at fixed frame rates, limiting their temporal resolution and energy efficiency. In this work we present an end-to-end pipeline that (i) converts real-world driving footage from the Joint Attention in Autonomous Driving (JAAD) dataset into synthetic dynamic vision sensor (DVS) event streams using the v2e simulator, (ii) augments training with the CARLA-simulated DVS sequences of the DVS-PedX dataset under both normal and adverse weather conditions, and (iii) trains a novel convolutional spiking neural network (Conv-SNN) with clip-consistent DVS augmentation to classify pedestrian crossing intent as binary: crossing or non-crossing. We detail all architectural decisions, the exact leaky-integrate-and-fire neuron dynamics with surrogate-gradient learning, the class-balanced loss formulation, JAAD oversampling at 6x, and a 70/15/15 stratified splitting protocol. The trained model achieves 95.83% accuracy and F1 = 0.9695 on the JAAD DVS test set, 97.79% accuracy and F1 = 0.9478 on normal CARLA DVS, and 94.78% accuracy and F1 = 0.8369 on adverse-weather CARLA DVS, all from a 1.07M-parameter architecture trained on CPU. Compared to prior frame-based approaches on JAAD, our method closes or surpasses the reported accuracy while operating natively on sparse temporal representations. We include a thorough analysis of the convergence behaviour across all 15 training epochs, domain transfer characteristics, and a quantitative comparison with representative related work.
Read the full abstract
Statistical Methodology
A tool for new Hybrid models: a swift algorithm to calculate the left inverse of a square root of a horizontally localized climatological ensemble covariance matrix
Computing matrix inverses (or, e.g., left inverses) using known algorithms is in general computationally prohibitive in high dimension, such as in an operational weather forecasting context. In this article, a swift, scalable algorithm to calculate the left inverse (through singular values decompositions techniques) of a square root of a horizontally localized ensemble covariance matrix is introduced. In this algorithm, calculations involving model variables located in different grid vertical columns can be performed separately. This is the aspect that makes this algorithm scalable by allowing for parallel computing. The algorithm is developed in the case when the number of ensemble members equals the number of model variables in each grid vertical column (this number is of order $10^3$ for current high-resolution models). The algorithm is tested in a synthetic experiment on a three-dimensional grid, and it is found to have very good accuracy. Additionally, a version of the algorithm outputting the left-multiplication of a generic vector by that above-mentioned left inverse is also introduced. For operational applications, if executed on a single processor, that version of the algorithm is already $9$ orders of magnitude faster than a traditional algorithm, and, with multiple, parallel processors, it scales until the number of processors reaches the order of $10^6$. In addition, an application of the algorithm is pointed out: performing the transform from model variables to an alternative set of variables, defined herein as standardized variables, being climatologically uncorrelated and each having unit climatological error variance. Furthermore, it is pointed out that the standardized variables allow constructing a new Hybrid covariance model, which is positive definite and climatologically unbiased, and it features a geographically dependent climatological error covariance matrix.
Read the full abstract
Geometric Topology
Signature Modules and the Dihedral Conjugation-Quandle Counting Invariant
We use reflection and rotation signatures of dihedral groups to develop a Smith normal form method for computing the conjugation quandle colorings across all dihedral groups. Let $L = L_1 \cup \cdots \cup L_\ell$ be a link with link components $L_1, \ldots, L_\ell$, and let $\operatorname{Conj}(D_n)$ be the conjugation quandle of the dihedral group $D_n$. We use the semidirect-product structure $D_n \cong \mathbb{Z}_n \rtimes \mathbb{Z}_2$ to separate the coloring into a component signature. This component signature tells us whether each link component is colored by rotations or reflections, and the exponents are assigned separately under modulus $n$. Once the component signature is determined, every crossing relation becomes an equation that determines an integer matrix $M_\tau$. The abelian group $A_\tau(L)$, which we call the signature module, is represented by an integer matrix $M_\tau$. We prove that the free rank and nonunit Smith normal form entries, or equivalently the isomorphism class of $A_\tau(L)$, are preserved under the Reidemeister moves. The complete family of signature modules also determines the entire family of counting invariants for all $D_n$. The all-reflection system is the same as the Fox-$n$ colorings, while the mixed rotation-reflection signatures contain information that gives a stronger invariant. We also show that each signature module is the $t_k = (-1)^{\tau_k}$ specialization of the multivariable Alexander module, and extend the coloring interpretation of the module to generalized dihedral groups $\operatorname{Dih}(B)$ for every abelian group $B$.
Read the full abstract
Graphics
SNAP3D: Physically Grounded 3D Parts for Assembly from a Single Image
Part-aware 3D asset generation enables applications such as editing, articulation, simulation, and fabrication, yet existing methods can generate visually complete individual parts without ensuring that they form a valid physical assembly. Consequently, generated neighboring parts may interpenetrate, lack valid connections, or collapse under gravity. We propose a physics-guided framework for improving single-image part-aware 3D generation with physically compatible geometry and stable connections. Our method resolves inter-part penetration, recovers a contact graph between neighboring parts, and introduces parameterized connectors at their contact surfaces. Using feedback from physical simulation, we refine connector placement, orientation, and dimensions to improve assembly stability while preserving the generated geometry. We further introduce a physics-based evaluation protocol that complements conventional geometric metrics by directly testing assembly validity and stability under gravity. Experiments comparing against multiple part-aware 3D generators show substantial improvements in physical realizability and stability while maintaining geometric quality. We additionally validate the resulting parts through 3D printing and real-world assembly.
Read the full abstract
Combinatorics
Alon's conjecture on intersecting families of graphs
Given a graph $F$, a family $\mathcal F$ of graphs on $[n]$ is \emph{$F$-intersecting} if $G\cap H$ contains a copy of $F$ for every $G,H\in\mathcal F$. We prove that there exists an absolute constant $\varepsilon>0$ such that every $P_4$-intersecting family $\mathcal F$ satisfies $|\mathcal F|\le\left(\frac12-\varepsilon\right)2^{\binom n2}$, which resolves a conjecture of Alon. Combined with Alon's reduction, this proves that a graph $F$ admits $F$-intersecting families of asymptotic density $1/2$ if and only if $F$ is a star forest. Multi-colored versions are also considered.
Read the full abstract
Computation and Language
Type Diversity Enables Transformers to Generalise Compositionally
Compositional generalisation has been divided into lexical and structural generalisation. Previous work has found that structural generalisation is harder than lexical for Transformers. We propose that this difference is not inherent to Transformers, but due to the high diversity of lexical types and low diversity of structural types in the specific datasets of these previous works. By type diversity we mean the number of different constructors of that type, instead of, for example, the specific word combinations that might populate the structure. To test this, we vary the amounts of type diversity of lexical and structural types in previously published datasets. We create linguistically diverse variants of the COGS and SLOG datasets using Grammatical Framework. We find that type diversity correlates with compositional generalisation equally in lexical and structural test cases, supporting our hypothesis. We note a contradiction with the proposition in previous work that compound divergence explains the difficulty in compositional generalisation tasks. We further investigate the effects of other dataset properties on compositional generalisation, such as the diversity of types other than the novel test structure, and surface properties of the logical semantics format.
Read the full abstract
Quantum Physics
Effective Quantization of Lossy Nonlinear Epsilon-Near-Zero Media
In epsilon-near-zero (ENZ) materials, the vanishing real linear permittivity results in the leading-order contribution to the displacement field being nonlinear in the electric field, making conventional canonical quantization approaches highly non-trivial. Existing treatments generally quantize the linear modes first and introduce nonlinear interactions subsequently, an ordering that becomes inadequate in the ENZ regime. Here, we provide, to our knowledge, the first effective single-excitation quantization in which the near-zero linear response, leading nonlinearity, and loss jointly determine the elementary excitation. Using a solvable microscopic atomic system as a theoretical scaffold, we find that this excitation is a polariton: a dressed quasiparticle with partly field and partly material excitations whose coefficients can be parameterized by macroscopic susceptibilities. While this work considers a leading-order $\chi^{(3)}$ nonlinearity, the framework could be generalized to include other nonlinear corrections, including a controllable second-order susceptibility $\chi^{(2)}$ and systematic higher-order contributions $\chi^{(n)}$, $n>2$. Our findings pave the way toward practical applications in quantum photonics, such as single-photon non-demolition detection, by leveraging the strong nonlinearities intrinsic to zero-index materials.
Read the full abstract
Number Theory
Serre's problem on statistics of Brauer symbols
We solve Serre's problem on the specialisation of Brauer group elements in all cases where the number of variables is sufficiently large.
Computation and Language
SAS: Simple Attention Sparsification via End-to-End Optimization of Context Ranking
Post-training attention sparsification reduces the quadratic cumulative attention cost of pretrained Transformers by selecting a small set of context units (tokens or blocks) for each query. Existing trainable methods usually use a lightweight selector to score context units, followed by hard Top-K selection that blocks gradients from the language modeling loss. Consequently, these methods commonly distill layer-wise dense attention distributions. Although this encourages the selector to rank context units by dense attention weights in the original model, the ranking is not directly aligned with their impact on predictions under a fixed attention budget (i.e., the number of attended context units per query), potentially wasting the limited budget on less useful units. To address this misalignment, we propose Simple Attention Sparsification (SAS), a gated sparse attention mechanism that optimizes context ranking end-to-end with the language modeling loss. The key idea is to inject the selector's continuous scores into attention logits during training, allowing the loss to update the selector through standard backpropagation. We identify several choices crucial for this simple design to work well in practice: placing the gate inside the attention softmax in log form, using normalized softmax gates to calibrate historical context against the always-retained current block, and preserving continuous selector scores so the model learns relative priorities rather than only hard selections. To support long-sequence training, we implement a memory-efficient Triton kernel that integrates SAS into FlashAttention-style computation. Across reasoning, long-context understanding, and agentic tasks, SAS consistently outperforms trainable sparse attention baselines across attention budgets, with especially large gains under tight budgets, demonstrating more effective context ranking for downstream tasks.
Read the full abstract
Cosmology and Nongalactic Astrophysics
Big Bang For Your Helium Buck
We study the implications of a recent measurement of the primordial helium fraction from the Large Binocular Telescope for cosmological inference from the cosmic microwave background. We show that LBT establishes the robustness of cosmological parameters to theoretical assumptions about big bang nucleosynthesis: its empirical calibration of the helium fraction enables constraints on cosmology and inflation that are agnostic to BBN but as precise as those that instead enforce standard BBN predictions. Future CMB surveys require at most a marginal improvement in precision over LBT to maximize their BBN-agnostic constraining power when the radiation density is free (and no more than a factor of two improvement across all cases we consider). We then apply the LBT measurement to a number of scenarios that feature new physics in BBN and the CMB. First, we constrain nonstandard radiation sectors (interacting and free-streaming) and search for evolution of the radiation abundance between nucleosynthesis and recombination. We then test models that alleviate the Hubble tension and the tension between CMB and baryon acoustic oscillation data, including self-interacting light relics and varying fundamental constants; LBT precludes most but not all of the models we consider via their effect on BBN. Finally, we use the LBT measurement as an indirect but independent test of the neutron lifetime anomaly, inferring values that are consistent with "bottle" experiments and $2.4 \sigma$ below "beam" experiments.
Read the full abstract
Materials Science
RS-CIDER: A non-local machine learning model for approximating screened hybrid functionals
Screened hybrid functionals such as HSE06 improve the description of band gaps, charge localization, and redox energetics relative to semilocal approximations, but their explicit Hartree-Fock exchange term is computationally costly for large, periodic systems, especially in plane-wave basis set calculations. Here we present RS-CIDER, a machine-learned non-local exchange functional that approximates the short-range Hartree-Fock exchange term in HSE06 by explicitly fitting both ground-state energies and single-particle energy levels. RS-CIDER combines scale-invariant semilocal and non-local density descriptors and can be evaluated self-consistently without explicitly applying the short-range Hartree-Fock exchange operator. RS-CIDER shows close agreement with HSE06 for molecular reaction energies and solid-state band gaps. Further tests across distinct materials show agreement between RS-CIDER and HSE06 for local magnetism, Cu-O phase competition, polaron localization, and neutral-defect energetics. For an Fe olivine, chemistry-specific fine-tuning recovers the HSE06 Li intercalation voltage. A timing benchmark shows that RS-CIDER reduces the measured per-SCF-step wall time by more than an order of magnitude relative to HSE06. Together, these molecular and solid-state results establish RS-CIDER as an efficient self-consistent machine-learned surrogate for HSE06.
Read the full abstract
Data Structures and Algorithms
Rank-1-perturbed trickledown theorems: Mixing time of Glauber dynamics for the Sherrington-Kirkpatrick model up to $\beta\leq \frac{1}{2}+\varepsilon$
We introduce a new family of trickledown theorems, a.k.a., local to global technique to bound the spectral gap of the Glauber dynamics for multi-state spin systems. In this technique instead of upper-bounding the influence matrix of a link of co-dimension 2 by $\lambda I$ (where $\lambda$ is the second eigenvalue of the link), we upper-bound the influence matrix after a carefully chosen rank-1 shift. The rank-1 shift allows for a significantly smaller upper-bound but it comes at the cost of bounding the average loss due to rank-1 perturbations. As an application we use this method to show that the natural Glauber dynamics mixes in polynomial time to generate samples from the Sherrington-Kirkpatrick model for $\beta\leq \tfrac{1}{2}+\varepsilon$, for an absolute constant $\varepsilon>0$. At the heart of the proof we manage to bound the loss due to rank-1 perturbations by averaging over all links of co-dimension 2.
Read the full abstract
High Energy Physics (Theory)
Mixed state entanglement measures for open string geometry
In the gauge/gravity framework flavor degrees of freedom are introduced to the gauge theory side through the insertion of probe flavor $D$ branes in the supergravity background. Scalar, vector or spinor fluctuations on theses flavor branes perceive neither the background supergravity metric nor the worldvolume-induced metric of the flavor branes. Indeed, they follow an effective metric called the open string metric (OSM). Through a proper choice of worldvolume gauge fields, one can induce a horizon structure in these OSMs. Studying holographic information-theoretic quantities in these kinds of geometries is equivalent to studying information-theoretic quantities in the flavor sector on the gauge theory side. In this article, we have studied various information-theoretic measures like mutual information, entanglement wedge cross-section and entanglement negativity for an open string geometry. All the calculations are done by considering strip-like parallel boundary subsystems for three- and four-dimensional OSMs. We have also measured the change in these mixed-state information-theoretic quantities from the relevant quantities in the pure AdS geometry due to the application of a background electric field.
Read the full abstract
Human-Computer Interaction
From Review to Reuse: How Post-Task Workflow Can Support Human-AI Agent Interaction
AI agents can automate tasks by turning a single natural-language request into a multi-step process spanning tools, files, and applications. Users are often left to judge that process from fragmented execution information and the final output. To make the completed process easier to understand, validate, and reuse, we investigate post-task workflows: editable, graph-based representations of an agent's completed execution. We first analyzed 10,803 public workflow templates from n8n to characterize real-world automation practice, then developed Trace2Flow, a research probe that translates agent execution traces into interactive post-task workflows. In a study, participants (N = 20) reviewed agent executions with prompt or agent errors. We found that post-task workflows improved their understanding and error detection over a prompt-only condition, and that validation succeeded mainly when users cross-checked across multiple evidence sources. For follow-up tasks, adapting the workflow matched adapting the prior prompt in success, time, and difficulty, and was often preferred.
Read the full abstract
Image and Video Processing
Recurrent Dynamic Range Extension
We present an approach to progressively extend the highlights of an image. Instead of reconstructing the full dynamic range of a complex scene directly, we learn a simpler task first: We extend the dynamic range of an input image by a single exposure value. Once this is mastered, we retrieve the full HDR image for the scene by executing our network recurrently, progressively increasing the dynamic range of the input. Our formulation is agnostic to the input dynamic range and targets a bounded output domain. This enables us to use widely available RAW images for the reconstruction task and adapt adversarial losses to construct realistic images. By incorporating Memory Replay for backpropagation, we can train our network recurrently over multiple inference stages and reduce reconstruction errors. As a consequence, our system reconstructs challenging long-tailed HDR scenes robustly and shows powerful recovery of bright light sources and highlights.
Read the full abstract
Artificial Intelligence
Rethinking Heterogeneous System Disaggregation for Subquadratic Attention
Frontier language models are more aggressively using subquadratic attention to reduce the memory footprint and compute requirements during inference while still delivering frontier accuracy. While existing systems make dense attention-centric disaggregated serving decisions, we show that disaggregating inference around the unique arithmetic intensity and memory footprint of subquadratic attention LLMs can achieve significant throughput and energy efficiency gains on emerging DRAM-based and SRAM-only heterogeneous systems. We introduce SQD (SubQuadratic Disaggregation), a fine-grained heterogeneous disaggregation scheme that splits decode by quadratic and subquadratic attention rather than by operator type, and that applies across subquadratic attention variants. For sparse attention LLMs, we disaggregate decode into top-k selection, which must index through the full KV, and top-k attention plus FFN, which have static memory footprints. For linear and sliding-window attention LLMs, we disaggregate decode into dense attention layers and subquadratic attention layers plus FFN. In an adjusted 8xB200 heterogeneous system proxy, we observe average tokens/J improvements of 53% on GLM 5.2, 31% on Nemotron 3 Ultra, and 56% on Gemma 4 31B over the strongest GPU-only baselines. In an analytical model of a Rubin plus LPX system with fixed power budgets, we observe 1.2x to 1.5x tighter achievable latencies and up to 3.6x higher throughput over the best baseline of attention-FFN disaggregation. Our experiments also reveal architectural insights on chip and interconnect provisioning for next-generation heterogeneous systems serving subquadratic attention.
Read the full abstract
Differential Geometry
The sharp $\sigma_k$-curvature inequality on locally conformally flat manifolds in quantitative form
Let $2\leq k<n/2$ and let $(M^n,[g])$ be a smooth, closed, connected, and locally conformally flat Riemannian manifold with a $k$-admissible metric in the conformal class $[g]$. We prove a stability result of the $\sigma_k$-curvature inequality on $M$, in the sense that if equality is almost satisfied for some conformal metric, then its conformal factor is close to a minimizer of the inequality. Closeness is measured quantitatively in terms of Sobolev norms of the conformal factor, namely with respect to the $W^{1,2}$- and the $W^{1,2k}$-norm. In the non-degenerate case, these norms come with optimal exponents $2$ and $2k$, respectively, whereas in general the exponents are $2+\gamma$ and $\max\{2k,2+\gamma\}$ for some $\gamma\geq 0$ originating from a Łojasiewicz inequality. This extends a previous result by Frank and the author from $k=2$ and the sphere to $2\leq k<n/2$ and the full class of manifolds originally considered by Viaclovsky. It also extends a previous result by Engelstein–Neumayer–Spolaor from $k=1$ to the setting of fully non-linear scalar curvatures.
Read the full abstract
Computational Complexity
A Dichotomy for Boolean Complex Holant Problems with Conjugate-Closed Signature Sets
We study Boolean Holant problems with complex-valued signature sets closed under conjugation. Such sets arise naturally in tensor-network expressions for classical strong simulation of quantum circuits. We prove a complexity dichotomy for such problems with an explicit tractability criterion. This extends the dichotomy for real-valued Holant problems, with the same four tractability conditions. Our proofs use Xia's projective binary group framework and quantum entanglement theory. The conjugate closure assumption precisely makes $k$-uniformity, directly applicable to the classification of Holant problems, by realizing reduced density matrices via Holant gadgets. We also use the classification of absolutely maximally entangled states to resolve a particular $6$-ary obstruction in our inductive proof of the \#P-hardness.
Read the full abstract
Number Theory
Eta-Quotient Representations for a Three Parameter Family of Modular Functions Associated with the Rogers-Ramanujan Continued Fraction
In 2021, Chern and Tang introduced two families of two-parameter modular functions associated with the Rogers-Ramanujan continued fraction. They established recurrence relations that express the members of these families in terms of eta quotients, and used these expressions to obtain dissection formulas. Motivated by their work, we construct a three-parameter extension and derive the corresponding recursive eta quotient representations. Our result also has applications to dissection formulas. In particular, it is used in separate work to obtain $5$-dissection formulas for overpartitions with restricted odd differences.
Read the full abstract
High Energy Physics (Phenomenology)
A Freeze-In Interpretation of the LZ High-Energy Nuclear Recoil Event
The recent LUX-ZEPLIN (LZ) experiment analysis reported a high energy nuclear recoil candidate at $E_R \simeq 248~{\rm keV}$, motivating interpretations in terms of inelastic dark matter. We study this event in a pseudo-Dirac fermion model with a vector boson mediator $Z'$, assuming a low-reheating temperature and dark matter production through freeze-in at stronger coupling. In contrast to the thermal freeze-out case, the reheating temperature provides an additional parameter controlling the relic abundance and breaks the one-to-one relation between the dark matter mass and direct-detection cross section. As a result, the LZ candidate can be reproduced over a continuous region of parameter space, including mass splitting below the thermal benchmark.
Read the full abstract
General Relativity and Quantum Cosmology
Tidal Love numbers of wormholes as black-hole mimickers
We study the dynamical scalar tidal Love numbers of wormholes that provide viable mimickers for black holes, focusing on thin-shell Schwarzschild and Damour-Solodukhin geometries. Using a matched near- and far-zone expansion, we determine their tidal response in the low-frequency regime. The presence of a long throat introduces an additional characteristic scale and naturally separates the modes into two classes. Super-throat modes probe the global wormhole geometry and are sensitive to both asymptotic regions, whereas sub-throat modes probe only one side of the wormhole and effectively perceive the throat as a black-hole horizon. We derive the scalar tidal Love numbers analytically for both classes of modes and show that their dissipative parts exhibit distinct low-frequency behavior, reflecting whether one or both potential barriers participate in the scattering process. We further find that, as the wormhole approaches the black-hole limit, the super-throat contribution becomes progressively negligible, while the sub-throat response smoothly reduces to that of a Schwarzschild black hole. These results demonstrate that the tidal response of wormholes depends crucially on whether the perturbation probes the global structure of the throat.
Read the full abstract
Signal Processing
Where to Defend? Layer-Wise Adversarial Training for Robust Transformer-Based Semantic Communications
Deep learning-based semantic communication (DeepSC), a Transformer-based encoder-decoder, achieves semantic fidelity over noisy channels but remains vulnerable to adversarial perturbations injected at multiple stages of the pipeline. We present a layer-wise robustness framework that compares fast gradient sign method (FGSM), projected gradient descent (PGD), and l2-normalized fast gradient method (FGM) defenses at the embedding output, and then uses PGD to analyze three attack and defense points: the embedding output, encoder output, and channel-encoder bottleneck. We evaluate reconstruction on Europarl and UK Hansard and sentiment classification on SST2 and YELP under additive white Gaussian noise (AWGN) and Rayleigh fading. A first-order damage budget, {epsilon} times the l1 norm of the clean-input loss gradient at each injection point, predicts the attack-severity ordering, and the transfer matrix reveals asymmetric defense transfer: the embedding defense transfers strongly to encoder attacks, whereas encoder defenses degrade robustness against upstream attacks. For reconstruction, encoder-point training yields the largest matched gain but fails severely under embedding attacks; joint embedding-plus-encoder training retains comparable gains under encoder attacks while mitigating this mismatch failure, and Rayleigh fading attenuates robustness gains and degradation. For classification, three of four defenses collapse to constant predictors; only the channel-encoder defense remains non-degenerate, suggesting a protective role for the 128D-to-16D bottleneck. At a signal-to-noise ratio (SNR) of 9 dB and perturbation budget {epsilon} = 0.3, the matched encoder defense recovers bilingual evaluation understudy (BLEU) from ~ 0.10 to ~ 0.64 on Europarl/AWGN, whereas the mismatched encoder defense yields a -0.444 BLEU change relative to the undefended baseline under an embedding attack.
Read the full abstract
Other Quantitative Biology
Making Models That Matter: How to Build Trustworthy and Useful Systems Biology Models
Computational models supporting mechanistic understanding of (complex) biological systems, systems behaviour prediction, and experimental design are becoming more and more embedded in research on complex biological systems. Reuse and refinement of models, rather than continuous reinvention, is becoming increasingly important as models' demands on computational infrastructure increase. However published models - despite the variety of efforts taken so far - are frequently difficult to reproduce or reuse, substantially limiting their scientific value. Here we address the requirements for model reusability in the light of the field-specific CURE framework (Credible, Understandable, Reproducible, Extensible) and the more general FAIR principles (Findable, Accessible, Interoperable, Reusable). Considering published guidance we identify broad agreement on requirements for findability, accessibility, and interoperability, but continued lack of clarity and consensus around reusability. Focusing on the scientific quality and usability of computational models we discuss six key practices underpinning model sharing and re-use. Mapping the FAIR and CURE principles onto the model lifecycle we propose ten recommendations for building and sharing systems biology models that are both FAIR- and CURE-compliant.
Read the full abstract
Number Theory
Quasipolynomial density bounds for $K$-point configurations in $\mathbb{Z}^d$
Let $d,K,N\in \mathbb{N}$ with $K\geq 3$ and $d\geq 4K+4$. Let $\Delta\subset \mathbb{Z}^d$ be the vertex set of a nondegenerate $(K-1)$-simplex, and let $A\subseteq[N]^d$ contain no nontrivial similar copy of $\Delta$. We prove that \[ |A|\ll_{\Delta,d} N^d\exp\!\left(-c_{\Delta,d}\sqrt{\log N}\right) \] improving upon a polylogarithmic bound due to Magyar. We perform a density increment argument using the circle method, and we introduce a “cut operator” method to decouple the weighted exponential sum over the system of quadratic forms describing the simplex. Our proof combines ideas from graph theory, functional analysis, and the geometry of numbers. In the process, we apply Finner's fractional form of Hölder's inequality, the analytic large sieve, and Kim's mean value formula for primitive lattice flags.
Read the full abstract
Artificial Intelligence
A Hybrid LSTM-XGBoost Framework for Multi-Horizon Stock Return Prediction Across Diversified Equity Portfolios
Accurate prediction of equity returns remains a major challenge in computational finance due to the non-stationary, nonlinear, and low signal-to-noise ratio nature of financial time series. This paper proposes a hybrid two-stage architecture that combines a long short-term memory (LSTM) network with an XGBoost gradient-boosted regressor for multi-horizon stock return prediction across a diversified panel of 14 U.S. equities spanning six industry sectors. The LSTM component, comprising two stacked layers with 64 hidden units, processes 60-day sliding windows of five sequential market features to produce 64-dimensional temporal embeddings that encode learned sequential market dynamics. These embeddings are concatenated with 14 hand-crafted technical indicators to form a 78-dimensional hybrid feature vector, which is subsequently passed to an XGBoost regressor tuned via 3-fold cross-validation grid search. The framework is trained on a multi-stock pooled corpus using strict chronological splits and per-stock MinMaxScaling to prevent look-ahead bias, and evaluated across four prediction horizons of 30, 90, 252, and 365 trading days. Experimental results demonstrate that the hybrid model achieves a test RMSE of 0.0949 on the 30-day horizon, roughly one-third that of the standalone LSTM baseline, while marginally matching or surpassing the XGBoost-Only baseline across the majority of stocks. Directional accuracy rises with horizon length, reaching 97.6% at 365 days; we show, however, that this largely tracks the high base rate of positive long-horizon returns in the sample, and we therefore benchmark directional accuracy against a naive always-positive predictor and treat the above-base-rate gap at short horizons as the more informative signal. A composite investment scoring framework derived from multi-horizon predictions is further proposed to support portfolio ranking and decision support.