Peer Revenue Equity in science

Papers

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

All of arXiv

Friday 28 August

Quantum Gases

Sound propagation in one-dimensional quantum droplets

Sound propagation in quantum droplets differs from that in conventional Bose-Einstein condensates (BECs) because of their self-bound nature and the role of quantum fluctuations. We investigate sound propagation in one-dimensional quantum droplets formed by a symmetric Bose-Bose mixture, focusing on finite-size and confinement effects. Using the extended Gross-Pitaevskii equation, we extract the sound velocity from the real-time propagation of localized density perturbations and compare it with the low-energy excitation spectrum. We find that, unlike in a conventional BEC, the sound velocity of a finite droplet is strongly affected by its density profile and quantum-pressure contribution. It decreases with increasing particle number as the droplet evolves from a Gaussian-like to a flat-top profile, approaching the bulk quantum-droplet value. In contrast, external harmonic confinement compresses the droplet and enhances the sound velocity, driving the system toward the acoustic behavior of a trapped BEC. Our results establish sound propagation as a sensitive probe of finite-size effects and the crossover between self-bound quantum droplets and conventional Bose gases, and suggest a feasible route for experimental observation in ultracold $^{39}$K droplets.
Read the full abstract

Artificial Intelligence

String: An Agentic OS Where Every App Is a Markdown File

LLM agents have become a new class of software user, but every surface they work through was designed for someone else. Pages are built for human eyes, which can skim and ignore; tool schemas for programs, which pay nothing to carry definitions they never call. An agent has neither luxury: it re-reads, and pays again for, everything it is shown on every turn. We present String, an open-source runtime that gives this user an interface of its own and treats the job as an operating-systems problem. Tool knowledge moves out of the agent's context and into a common layer that renders it back one view at a time as Markdown. A single SFMD (String-Flavored Markdown) document declares an application's views, typed actions, navigation, and credentials, and the runtime handles discovery, validation, execution, state, and secrets behind two core verbs: /open to see and /act to do. Web and app turn out to be two renderings of one architecture: an SFMD site serves styled HTML to browsers and the raw document to agents, so one grammar reaches apps, files, shells, and the web, even legacy HTML, with no per-site integration. Views stay partial by design, and the staging is causal: disclosing one tier of detail a single turn too early costs up to 23 accuracy points, while proper staging drops wrong-action selection from 28% to 2%. Privilege follows provenance: a remote page may call HTTP but never the shell, and caller-supplied text never expands a stored secret. On an 87-task benchmark that pairs each task with curated skills, operationalizing those procedures as on-demand String apps yields comparable aggregate success across six models from frontier to small (+1.3pp) while using 33.5% fewer tokens among completed episodes, and the resident interface stays a constant 53 tokens at any catalog size. We report the design, the evaluation, and what three months of production use taught us.
Read the full abstract

Sound

Klangfarbenakkord and Klangfarbenharmonien Metric Space Models for Music on Informational Geometry 1

This paper deals with the introduction of "geometric harmony", a discipline that explicitly addresses the spectral characteristics of musical gamut. The framework of Western music, from Renaissance to the present, represents sound in terms of "pitch"-as is evident from its five-line staff notation system-and employs the fundamental frequency as its representative value, 440 Hz, etc. In this paper, by taking the timbres of specific individual instruments as elements and examining the Wasserstein distance between two voices, and Wasserstein deviations between three or more voices, we demonstrate that it is possible to expand the system whilst retaining the entire framework of conventional music theory. At the same time, as an example of practical utility in ensemble playing, we provide a detailed account of the two-voices affinity of the "Throat G" on clarinet, a note known for its fragility in ensemble contexts.
Read the full abstract

Chemical Physics

Model-free Reconstruction of Molecular Energy Levels by Broadband Kilohertz-accurate Cavity-enhanced Spectroscopy

Assigning the lines of high-resolution molecular spectra to quantum states requires a Hamiltonian model and substantial expert intervention, so the spectra of larger molecules accumulate vast numbers of unassigned lines. Here we show that molecular energy levels can instead be reconstructed directly from the raw, unassigned transition frequencies, using graph theory alone with no model and no prior assignment. Our inverse graph construction exploits recurring frequency differences and four-cycle closures to assemble an energy-level network at kilohertz precision. The dense, broadband spectra this requires are produced by a cavity-enhanced spectrometer (SCALS) that scans continuously across tens of terahertz at kilohertz accuracy, combining broadband coverage, high sensitivity, and high precision in a single automated instrument. Applied to the water absorption spectrum in the range of 1537–1605~nm, 686 Lamb dips of water were obtained without assignments, and the method reconstructs 158 energy levels that are mostly two orders of magnitude more precise than the corresponding literature values. By removing the assignment barrier, this approach opens a route to exploratory precision spectroscopy of polyatomic molecules without a priori knowledge of transition frequencies.
Read the full abstract

All authors S. Vasilchenko, A.-W. Liu, C.-X. Zuo, Y.-Q. Cheng, Z.-T. Zhang, W.-T. Wang, Y. R. Sun, Y. Tan, S.-M. Hu

Representation Theory

Lusztig-Vogan categories of equal rank 2

Lusztig-Vogan categories are categorifications of the principal block of the Lusztig-Vogan module over the Hecke algebra, which captures information about characters of irreducible admissible representations of a real reductive group. Lusztig-Vogan categories can be constructed as module categories over Soergel bimodules. In this paper, we describe the structure of the rank 2 Lusztig-Vogan categories corresponding to equal rank real groups. More precisely, we classify indecomposable objects and describe the action of generating Soergel bimodules, recovering the $W$-graph of the underlying Lusztig-Vogan module. We also provide an algorithm which completes this procedure for arbitrary finite rank Lusztig-Vogan categories, including those which do not correspond to a real reductive group.
Read the full abstract

Fluid Dynamics

Validation of quasi-two-dimensional model of convection in a transverse magnetic field

In this study we examine applicability of the quasi-two-dimensional (Q2D) model of natural convection in a laterally heated rectangular box subject to a transverse magnetic field. For this purpose, results obtained using the Q2D model are compared with fully three-dimensional calculations for rectangular boxes with a square cross-section and horizontally elongated ones with the length to height ratio 8. The width to height ratio is varied from 1 to 10. It is argued that the Q2D model is valid under certain conditions, which include perfect thermal and electrical insulation of the spanwise boundaries and reflection symmetry with respect to the spanwise midplane. It is shown that under these conditions, the Q2D model yields quantitatively correct results for boxes with square cross-section, as well as for horizontally elongated cavities, independently on the width to height ratio. Three-dimensional results obtained for boxes with perfectly conducting walls exhibit asymptotics at gradually increasing width ratio, which, however, does not agree with the predictions of an amended Q2D model. It is shown also that when is applicable, the Q2D model can correctly predict steady-oscillatory transition of three-dimensional flows. It is proposed to seek for extensions of the Q2D model for cavities with perfectly conducting spanwise boundaries.
Read the full abstract

Numerical Analysis

Transport-Matched Penalties for Diffusion Synthetic Acceleration of Polytopic Discontinuous Galerkin Discretisations

Diffusion synthetic acceleration is most effective when its diffusion correction reflects the transport discretisation that generates the iteration error. We develop this principle for high-order upwind discontinuous Galerkin discretisations of discrete-ordinates transport on polytopic meshes. From the discrete transport sweep, we derive the exact scalar correction that removes the source-iteration scalar error in one step. We prove that the associated scalar response is positive and self-adjoint, obtain an exact expression for the source-iteration convergence factor, and quantify the additional damping produced by vacuum leakage. Using the exact correction as a reference, we construct a transport-matched modified interior penalty correction whose boundary terms are inherited directly from homogeneous vacuum inflow. In the optically thick regime, the resulting MIP form approximates the exact correction with relative error proportional to the effective cell Knudsen number. This gives a strict acceleration of source iteration, with bounds uniform in mesh size, polynomial degree, and element face count for admissible polytopic meshes. Numerical experiments on Cartesian and centroidal Voronoi meshes confirm the predicted convergence and correction-operator scaling.
Read the full abstract

All authors Ansar Calloo, Matthew Evans, Francois Madiot, Tristan Pryer

Cryptography and Security

Compared to What? A Human-Anchored Security Benchmark for LLM-Generated Infrastructure-as-Code

Large language models increasingly author Infrastructure-as-Code (IaC), where one insecure default is provisioned straight into production. Prior evaluations report vulnerability counts for models only, and so cannot say whether models are worse than the engineers they assist. We present GenIaC-SecBench: 100 deployment scenarios across 12 model configurations from six vendors, open and closed weights, yielding 1,196 artifacts scanned by three policy engines (Checkov, Trivy, KICS) at complete coverage. Crucially we scan 634 human-authored IaC templates with the identical toolchain, giving the first size-matched human security baseline for this task. Vulnerability density is strongly inverse to artifact size (Spearman $\rho=-0.55$, $p<10^{-77}$), so unmatched comparisons measure size, not security. Size-matched, every configuration exceeds the human baseline at $3.21\times$ to $3.87\times$, and the gap widens as tasks get simpler ($4.9\times$ at one resource, $1.4\times$ at twenty or more). A majority of scenarios prescribe a security state rather than specifying function alone, so we stratify by prompt class: pooled the gap is $3.50\times$, and excluding every scenario that explicitly requests an insecure configuration still leaves all configurations above baseline ($2.4\times$ to $4.2\times$). The corpus cannot isolate unprompted default posture, and we say so. Decomposing "reasoning" into standard generation, prompted chain-of-thought, and vendor extended-thinking APIs, extended thinking beats prompted CoT ($-12.0\%$, $p=0.0013$) while prompted CoT alone is indistinguishable from standard ($-1.3\%$, n.s.); it consumes under $1\%$ of the output budget, bounding the effect. Two negative results: more deployable models are not more vulnerable ($r=0.158$, $p=0.625$), and complete-case Friedman is uncomputable here, motivating Skillings-Mack. All code and data are released.
Read the full abstract

2608.28021 · 28 Aug · 14 pages, 12 figures, 10 tables. v2 corrects the prompt characterization: v1 called the scenarios functional-only, but a majority prescribe a security state, so those findings partly measure instruction-following, not unprompted defaults. Found by an independent audit (Chauhan, IaC-Guard-V, QRS 2026), reproduced exactly. Results now stratified by prompt class; pooled gap unchanged at 3.50x

Computer Vision

3D-USE: From Image-Level to Scene-Level Underwater Enhancement

Underwater scene reconstruction from degraded views can preserve visual degradation. Since paired enhanced 3D references are hard to obtain in real ocean scenes, underwater 3D methods largely rely on physics-based restoration. However, separating object radiance from medium effects is ill-posed and may leave residual artifacts. In contrast, we formulate underwater scene enhancement as learning the desired appearance of the scene and propose a two-stage framework called 3D-USE. Stage 1 builds a medium-aware Gaussian scene with MediumRBF, sharing medium estimates across nearby views and establishing stable geometry and visibility. Using a prior learned from external 2D enhancement pairs, Stage 2 applies Appearance Transition Consensus (ATC) to calibrate and aggregate view-wise appearance changes, defining the desired scene-level appearance transformation. Finally, an Underwater Bilateral Appearance Field (U-BAF) jointly adjusts object radiance and medium appearance to produce coherent enhanced appearance across viewpoints. Experiments show that 3D-USE improves novel-view enhancement and cross-view consistency while maintaining competitive reconstruction quality. The project page and code are available at https://bilityniu.github.io/3D-USE/.
Read the full abstract

All authors Jieyu Yuan, Yuanlin Zhang, Jihong Li, Chunle Guo, Huimin Lu, Chongyi Li

Earth and Planetary Astrophysics

Cooling-regulated gas accretion onto gap-opening planets

Gas accretion onto forming planets controls the final masses of giant planets and provides observable signatures of ongoing formation. How this process depends on the cooling properties of these newly attracted gas remains poorly constrained. We present long-term, three-dimensional global hydrodynamical simulations to quantify gas accretion onto gap-opening planets in the mass range between 1 and 3 Jupiter masses. We systematically vary the cooling time, $\beta$, from near-isothermal ($\beta=10^{-2}$ in units of orbital time) to near-adiabatic ($\beta=10^{2}$), and follow the evolution until a quasi-steady state is reached. Our simulations show that the gas accretion rate decreases monotonically with increasing $\beta$, as $\dot{M}_{\rm acc}\propto\beta^{-0.18}$, reaching values at $\beta=10^2$ that are approximately an order of magnitude lower than locally isothermal predictions, largely independent of planet mass. The reduction in accretion is traced to thermodynamic restructuring of the circumplanetary region: inefficient cooling weakens shocks, narrows the accretion bands feeding the circumplanetary disk. Our results imply that thermodynamic effects should be taken into account when interpreting observed accretion rates of young planets, and may introduce systematic uncertainties in commonly used locally isothermal assumptions.
Read the full abstract

All authors Ayumu Kuwahara, Michiel Lambrechts, Minghao Zhang, Ruobing Dong, Hideko Nomura

Computation and Language

Twin Worlds: Equivariance-Based Abstention for Evidence-Grounded Reasoning

Knowledge-intensive reasoning requires Large Language Models (LLMs) to ground answers in provided evidence. When evidence is insufficient, it is desirable that models abstain rather than confidently generating unsupported answers. Existing abstention methods rely on uncertainty estimation or evidence sufficiency checks, but neither tests whether the reasoning process for generation, driven by the interaction of provided evidence and the model's internal memory parameters, is actually grounded in the evidence. A key contributing factor is that entity mentions in context activate memorised associations, causing models to generate plausible responses ungrounded in evidence. We propose Twin Worlds (TW), a framework for improving reliability in knowledge-intensive reasoning through equivariance-based abstention: unlike invariance, which requires outputs to remain unchanged, equivariance requires outputs to transform correspondingly under entity substitutions. A model grounded in the evidence should produce answers that shift consistently when entities are substituted while their relations are preserved. TW constructs multiple worlds via typed substitutions of the original input that preserve relational structure while reducing parametric priors, and uses equivariance violations as an abstention signal. Across four benchmarks and three model backbones, TW identifies when answers are not reliably grounded in the provided evidence and outperforms uncertainty- and sufficiency-based baselines.
Read the full abstract

All authors Vy Nguyen, Ziqi Xu, Jeffrey Chan, Estrid He, Feng Xia, Renqiang Luo, Erik Cambria, Xiuzhen Zhang

Systems and Control

Securing Cooperative Sensing in UAV Swarms Against Conformity-Driven Byzantine Attacks

In integrated sensing and communication (ISAC) enabled 6G unmanned aerial vehicle (UAV) swarm networks, the widely adopted imitation-based conformity cooperation mechanism can be exploited by Byzantine attackers to fabricate false consensus, causing the effective error probability of normal UAVs to evolve dynamically and far exceed their inherent sensing errors, which invalidates conventional fusion methods built on the independence assumption. This paper proposes a conformity-aware Byzantine-resilient fusion framework that couples evolutionary game theory with maximum a posteriori (MAP) estimation. First, the strategy updates of normal UAVs are characterized by bounded-rational opinion dynamics, and the evolution dynamics of the misinformation ratio together with its evolutionarily stable state (ESS) are derived under death birth updating. Three theoretical results are then established: under heterogeneous per-node sensing errors, the zeroth-order ESS depends on the error distribution only through its mean; a closed-form first-order weak-selection correction to the ESS is obtained, together with an exact mean-field fixed point valid for arbitrary selection intensity; and it is revealed that swarm level misinformation can overwhelm the majority if and only if the attack probability exceeds one half, with this threshold independent of both the sensing error and the malicious ratio. Embedding the predicted error dynamics into a per-node MAP rule, the resulting fusion mechanism achieves nearly 100% situation-inference accuracy under different network topologies, attack intensities, network scales, and sensing-error distributions, and maintains accuracy above 99% under +-20% parameter mismatch. In contrast, majority voting, reputation weighting, and independent fusion collapse completely once the majority-flip threshold is crossed.
Read the full abstract

All authors Ruixing Ren, Junhui Zhao, Qiuping Li, He Fang, Jiamin Li, Dongming Wang

Cryptography and Security

The Impact of Magma: A Ground-Truth Fuzzing Benchmark

Magma is an open-source and ground-truth fuzzing benchmark that enables uniform fuzzer evaluation and comparison. Magma was originally released with a research paper published at ACM SIGMETRICS 2021. This short paper explains the motivation, the design, and the impact of Magma, with a description of extensions to the original benchmark.
Read the full abstract

All authors Ahmad Hazimeh, Adrian Herrera, Srividya Subramanian, Thaqiya Aman, Sara Vaccino, Qiang Liu, Mathias Payer

Optimization and Control

The AHI family of sum of squares polynomials

We introduce a new family of non-negative polynomials, constructed via the arithmetic harmonic inequality, called AHI polynomials. We derive explicit algebraic conditions for this family and prove that, for AHI polynomials, the cone of non-negative polynomials coincides with the cone of sum of squares (SOS) polynomials. We then study their convexity, showing that although AHI polynomials are generally nonconvex, certain monomial substructures are SOS convex. We further locate the family precisely among the standard nonnegativities certificates; every AHI polynomial is simultaneously SOS and a sum of non negative circuit polynomials (SONC), and the containment in the intersection of these two cones is strict. By closing this family under multiplication, we obtain a cone Pi AHI that is, by construction, still SOS, yet we prove that it lies outside both the SONC cone and the smaller SDSOS cone. Moreover, membership in this cone admits a closed form certificate that does not require solving any semidefinite programs. Finally, we demonstrate the usefulness of these structures in optimization, numerical experiments indicate that exploiting AHI sparsity yields a computation time over 300 times faster than dense SOS relaxations and enables solving high degree polynomial optimization problems (up to degree 40) that standard methods cannot handle due to computational limits, and a factorized hierarchy for Pi AHI decomposes products into independent small subproblems that generic sparsity techniques do not detect.
Read the full abstract

Number Theory

Drinfeld modular forms of higher rank from a lattice-oriented point of view

A space $\mathcal{L}_N^r$ of Drinfeld modules of rank $r \geq 1$ with level structure, or equivalently lattices of rank $r$ with level structure, is introduced, and its irreducible components and group actions on it are investigated. A metric is defined on this space, its completion $\overleftarrow{\mathcal{L}_{N}^{r}}$ is established and the aforementioned group actions are extended to the completion. A decomposition of the completion into multiple smaller spaces $\mathcal{L}_N^s$ is proven. Drinfeld modular forms are defined as homogeneous holomorphic functions on $\mathcal{L}_{N}^{r}$ which are continuous on the completion $\overleftarrow{\mathcal{L}_{N}^{r}}$, and the group actions above are extended to actions on the spaces of modular forms. Finally, the modular forms defined here are compared with those of Basson, Breuer, and Pink and with those of Gekeler, and it is shown that the cusp forms (those which are zero on the boundary) coincide.
Read the full abstract

Combinatorics

A product theorem for $r$-cross intersecting families of subspaces

Let $V$ be an $n$-dimensional vector space over a finite field of order $q$. Let $r\geq 3$, $(r-1)n\geq rk$ and let $\mathcal F_1,\ldots,\mathcal F_r\subset \genfrac{[}{]}{0pt}{}{V}{k}$, where $\genfrac{[}{]}{0pt}{}{V}{k}$ denotes the set of $k$-dimensional subspaces of $V$. Suppose that $F_1\cap\cdots\cap F_r\neq\{0\}$ holds for all $F_i\in\mathcal F_i$, $1\leq i\leq r$. Then we show that $\prod_{i=1}^r|\mathcal F_i|\leq\genfrac{[}{]}{0pt}{}{n-1}{k-1}$, provided $n-k$ is sufficiently large for fixed $q$ and $r$. Moreover, equality holds if and only if there is a common line $L$ such that every family $\mathcal F_i$ consists of all $k$-dimensional subspaces containing the line $L$. One of the main tools of the proof is a junta theorem concerning intersecting linear maps obtained by Ellis, Kindler, and Lifshitz.
Read the full abstract

Statistical Mechanics

Correlated disorder versus correlated noise: Ordering in active systems

Can quenched disorder generate ordering in driven systems? Using a recently proposed hydrodynamic model, we show that sufficiently long-ranged quenched disorder can induce long-range order in two-dimensional (2D) nonreciprocal XY systems, even when the clean system exhibits only short-range order at finite noise. Active surfaces tangentially advected by quenched velocities, governed by the same hydrodynamic equation, become statistically flat with super- or subdiffusive relaxation. In three dimensions (3D), quenched disorder combined with nonreciprocity produces a novel transition between strong-coupling and asymptotically noninteracting regimes, supporting either long- or short-range order. In both 2D and 3D, the exponents are nonuniversal, which vary continuously with the degree of transversality of the quenched disorder. The degree of transversality of the quenched disorder can be tuned to induce transitions in the model for fixed disorder and noise variances.
Read the full abstract

Artificial Intelligence

Coverage, Not Credit: Failure-Credit Routing of Zeroth-Order Perturbation Budgets Does Not Improve On-Pool Sample Efficiency for LLM Agents

Trajectory-level credit assignment can localize which module of a tool-using LLM agent causes failures using only verifiable signals. We ask whether such failure credit should route a fixed zeroth-order/evolution-strategies (ZO/ES) perturbation budget. Across a synthetic environment and frozen Qwen2.5-1.5B/3B and SmolLM2-1.7B agents, three task families, six allocation schemes, a credit-noise sweep, paired seeds, and exact sign-flip tests, we find no statistically detectable improvement over uniform allocation in any on-pool comparison (no gain of at least 2 percentage points). The joint soft-plus-sigma scheme is equivalent to uniform within a +/- 0.02 AUC margin on 1.5B and 3B; concentrating the full budget on the credit argmax is marginally equivalent on 1.5B, where that module is the verified bottleneck, and significantly worse on 3B. Inverse-propensity debiasing does not rescue routing, and misrouting costs up to -0.074 AUC in-house and -0.118 end-to-end on the BFCL-derived family. Across six fixed-step schedules, loss is linear in bottleneck starvation rate (R^2 = 0.94, descriptive), and a preregistered credit-free coverage floor removes detected harm. Matched-budget burst and step-compensating catch-up schedules are consistent with harm arising from insufficient cumulative parameter movement rather than update frequency. Our primary estimand is optimization efficiency on a fixed task pool. On unseen BFCL functions, the study's one exception is that soft routing exceeds uniform on held-out endpoints (+0.047, p = 0.031, n = 6). A plausible but untested reading is that routing-favored caller improvements transfer while uniform's on-pool gains reflect a synthesizer behavior specific to our harness. We report this exception explicitly and document three failure modes that can silently invalidate ZO/ES experiments on frozen LLMs.
Read the full abstract

Machine Learning

When Can Conditional Flow Matching Replace Pointwise Negative Log-Likelihood?

Flow matching enables likelihood-free training, yet alignment methods increasingly reuse conditional flow matching (CFM) losses as endpoint negative log-likelihoods (NLLs) and their old/new differences as log-likelihood ratios. We characterize when these substitutions are valid. For linear Gaussian paths, we exactly decompose endpoint NLL into entropy, a weighted CFM objective, an interior velocity–score residual, and a boundary residual. Thus CFM-only estimates and differences are exact only when the corresponding residuals cancel. At the off-policy population optimum, ordinary CFM is not generally a pointwise NLL estimator, whereas \(w_{\mathrm{sc}}(t)=(1-t)/t\) removes the interior residual; this positive result does not extend generally to training or on-policy alignment. On-policy log-ratios can remain biased even for identical endpoint laws or after surrogate optimization. Experiments across dimensions, distributions, and geometries support these conclusions and the mechanisms that make inexact ratios useful. **More broadly, the decomposition provides a theoretical basis for adapting likelihood-based LLM methods to flow matching, while distinguishing exact substitutions from controlled surrogates.**
Read the full abstract

Computation and Language

Beyond Global Scalars: Synergizing Token-Level Statistics and Deep Semantics for Adversarial AIGC Text Detection

The rapid evolution of large language models necessitates robust machine-generated text detection. Existing paradigms typically follow two isolated tracks. Training-free methods rely on global statistical scalars such as perplexity, while training-based methods utilize semantic hidden states. Both approaches exhibit fundamental vulnerabilities in adversarial scenarios. Global scalars act as lossy compressions that obscure local probabilistic burstiness in interleaved texts, whereas pure semantic models overfit to specific fingerprints and remain susceptible to spoofing. To expose these flaws, we introduce MOSAIC, a comprehensive adversarial benchmark comprising 16000 samples across a full-granularity attack spectrum. To address these challenges, we propose NeuroStat, an end-to-end framework bridging the statistical and semantic gap. NeuroStat captures uncompressed token-level probabilistic logits alongside deep semantic hidden states from a single causal language model backbone. We fuse these heterogeneous signals through Macro-State Residual Modulation, which adaptively calibrates local convolutional features using global uncertainty indicators. Orthogonal and contrastive losses further ensure the learning of complementary representations. Extensive experiments demonstrate that NeuroStat maintains exceptional robustness on MOSAIC compared to the severe degradation of state-of-the-art methods, establishing a new standard for adversarial text detection. Code and the MOSAIC benchmark are available at https://github.com/TencentBAC/NeuroStat.
Read the full abstract

All authors Peiming Li, Yifan Wang, Zhiyuan Hu, Shiyu Li, Zheng Wei, Yang Tang

Computer Vision

Visual Token Coding for Video Multimodal Large Language Models

In this paper, we propose a new token compression paradigm for video Multimodal Large Language Models (MLLMs), termed Visual Token Coding ($VTC$). Inspired by classical video coding principles, e.g., HEVC, VTC performs structured compression by predicting the I/P frames of a video and measuring their frame-wise residuals to estimate token redundancy. Based on this base paradigm, we further extend VTC with a set of dynamic budgeting designs towards more cost-effective token coding, termed $VTC_{dy}$. To validate VTC and $VTC_{dy}$, we apply them to three MLLMs and conduct experiments on a set of video understanding benchmarks. The experimental results show that as a base paradigm, VTC can achieve an average performance retention of 98.4% on Qwen3-VL under a 50% token compression ratio, and the enhanced $VTC_{dy}$ can further improve the retention to 100.3%, outperforming a set of compared methods. Moreover, as a plug-and-play design, VTC requires no additional tuning of MLLMs for token coding. Our code is given in the supplementary materials.
Read the full abstract

All authors Chenxin Fang, Tao Chen, JunChao You, Jun Peng, Yiyi Zhou, Rongrong Ji

Machine Learning

Exact Risk Ratios for Weighted Data Selection in Linear Regression

How much data must a fixed learner retain? Hanneke, Moran, Shlimovich and Yehudayoff (COLT 2025) posed this question for linear regression with the minimum-norm empirical risk minimizer. A selector sees a finite dataset $D\subseteq R^d\times R$, keeps at most $n$ examples with nonnegative weights, and $F_w(d,n)$ is the worst-case ratio between the full-data loss of the trained predictor and the optimal loss. The value is $\infty$ for $n<d$, $d+1$ at $n=d$ and $1$ for $n\ge2d$, and the regime $d<n<2d$ was left open. We settle several cases. For every $d$ we prove $F_w(d,2d-1)=1+1/d$, which confirms a claim stated without proof in the original note. We also prove $F_w(3,4)=5/3$, $F_w(4,5)=2$ and $F_w(4,6)=3/2$, the three smallest cells not covered by that formula. For every intermediate budget $n=d+k$ we prove the lower bound $F_w(d,d+k)\ge1+\Gamma_{d,k}$, where $\Gamma_{d,k}$ is an explicit harmonic quantity over balanced partitions of $d$. This bound is the exact minimax value on the class of datasets whose whitened systems split into orthogonal circuit blocks. All proved values equal $1+\Gamma_{d,k}$, and we conjecture that this holds throughout the open regime. Our upper bounds combine a rigidity theorem for positive spanning configurations of loss gradients with normal forms of the small positive bases in $R^3$ and $R^4$. These forms are special cases of the classification of Cornaz, Kerleau and Royer; we also control all gradients outside the basis. A dimension-free extremal-basis argument converts sign-cone geometry into selections of $d+1$ points. Explicit counterexamples rule out several shorter routes. Every upper bound is constructive, with selection procedures polynomial in the number of points for fixed dimension. Every numerical claim about a specific instance is an exact rational or algebraic identity, recomputed in exact arithmetic in the supplementary material.
Read the full abstract

Rings and Algebras

Commuting graphs of Okubo algebras

Commuting graphs of Okubo algebras are considered, and the problem of their connectivity is studied. The commuting graph of a pseudo-octonion algebra $P_8(\mathbb{F})$ over a field $\mathbb{F}$, $\mathrm{char} \, \mathbb{F} \neq 3$, that contains a primitive cubic root of unity is shown to be isomorphic to the commuting graph of the matrix algebra $M_3(\mathbb{F})$. As a consequence, if the field $\mathbb{F}$ is algebraically closed, then the diameter of the commuting graph for the unique Okubo algebra over $\mathbb{F}$ equals $4$. It is shown that the commuting graph of the real division Okubo algebra is connected, and its diameter also equals $4$. The proof of this result relies on the fact that, given any two idempotents in an arbitrary Okubo algebra, the intersection of their centralizers is always nonzero.
Read the full abstract

Chemical Physics

Projected q-Shells for Nested Gausslet Bases

Electronic-structure calculations require a finite representation of both the electronic wave functions and their Coulomb interaction. Conventional compact basis sets lead to a four-index interaction. Gausslet basis sets are one of a very small number of approaches that allow a much smaller two-index interaction, similar to a real-space grid, but with far fewer points. The continuing challenge in their development is reducing further the number of functions needed while maintaining high accuracy. Here we introduce projected q-shells (PQS), an improvement to the nested Cartesian gausslets of White and Lindsey. Their standard nesting construction divides a joining shell into disjoint faces, edges, and corners, reducing its polynomial completeness from order q to order q-2. PQS approximately restores the two lost orders. In terms of completeness, we prove that for an ideal, undistorted parent, the span of the PQS basis is equivalent to an overcomplete basis of overlapping patches. Coordinate distortions, which are always used in practice, harm this equivalence only slightly. The moment properties that allow two-index interactions also become inexact, even without distortion, but we show that the improvement in order more than makes up for this loss. For similar basis sizes, PQS gives smaller one-electron and interaction errors on small molecules, such as H2+, H2, and C2, than standard nesting. The improvement is largest for more compact bases with smaller q.
Read the full abstract

Probability

Hausdorff Dimension of the Set of Extreme Points of a Random Countable Stable Zonotope

Let $d\geq 2$, $0<\alpha<1$, and let $\Gamma_k$ be the successive arrival times of a standard Poisson process on $(0,\infty)$. Given independent uniform directions $\varepsilon_k\in S^{d-1}$, independent of $(\Gamma_k)$, we consider the random countable stable zonotope $Z_\alpha=\bigoplus_{k=1}^{\infty}\Gamma_k^{-1/\alpha}[0,\varepsilon_k]$. For its set of extreme points $\operatorname{ext} Z_\alpha$, we prove that almost surely $\dim_H \operatorname{ext} Z_\alpha=(d-1)\alpha$, and that the critical Hausdorff measure $\mathcal H^{(d-1)\alpha}(\operatorname{ext} Z_\alpha)$ is almost surely finite. The lower bound follows from the tangential non-degeneracy of the stable increments of the parametrizing field and Frostman's energy criterion. For the upper bound we construct an adaptive covering: at each scale the Poisson jumps are split into large and small ones, the large jumps determine a finite hyperplane arrangement, and the sum of the small jumps controls the diameters of the images of its cells.
Read the full abstract

Machine Learning

A Method for Layer Bit-Width Allocation in LLM Quantization via Performance Maximization Under a Quality-Degradation Constraint

This paper proposes a layer bit allocation method for Gemma-3-1B, formulating the problem as performance maximization (latency decrease) given a degradation budget constraint (allowable level of generation quality loss). This approach is different from time- and resource-consuming uniform layer quantization methods that are used in the literature (like GPTQ or AWQ) or allocation methods without proven performance-accelerating effect (like MixLLM or TorchAO). The layer sensitivity profile resulting from our prior work SA-PTQ is applied using the activation pass-through mode inside TensorRT-LLM. For each layer precision is determined individually in blocks, according to a grouping introduced in the prior step (5+5, 10+10, all26), differentiating the contribution of FFN, Attention, and lm_head to the overall speedup. The clock speed was measured for 13 W8A8 variants on an RTX 5090. We find that for FFN and lm_head the time cost of quantization/dequantization is compensated for by the use of integer arithmetic, while for short context lengths, the opposite holds true for Attention: an additional step of quantization slows execution down. We propose a manual implementation of SmoothQuant for TensorRT-LLM which was necessary due to export failures, unavailable for lm_head. The best solution found under joint consideration of all three criteria with minimal degradation was FFN 5+5 with lm_head, providing an 11.0% reduction in latency with negligible quality loss (98.90% Top-1 agreement, +0.85% perplexity degradation). With acceptable quality loss for FFN all26 + lm_head, a speedup up to 19.1% was found possible. We suggest further optimizations: fused attention kernels in INT8, KV-cache quantization, using FP8 instead of INT8 and partial Attention quantization analogous to FFN.
Read the full abstract

Optimization and Control

On Solving Continuous-Discrete Projection Filters via Sum-of-Squares Relaxation

The continuous-discrete projection filter offers a rigorous framework to approximate the solution of the nonlinear state estimation problems. However, it suffers from numerical instability during the prediction phase when integration errors force the natural parameters outside their admissible domain. To address this issue, we introduce the sum-of-squares (SOS) relaxation to constrain the evolution of the natural parameters within the admissible domain. By parameterizing the underlying SOS matrix using the log-Cholesky map, we derive a projected ordinary differential equation (ODE) that inherently preserves the necessary positivity constraints without requiring the computationally expensive online optimization checks associated with previous semi-infinite programming approach. We provide a theoretical derivation of this positivity-preserving propagation scheme and present the explicit SOS-relaxed evolution equations for the Gaussian case.
Read the full abstract

Human-Computer Interaction

FocusGen: Expanding Visual Design Exploration with a Simulated Focus Group of Persona Agents

Creative professionals rarely design for themselves–they design for audiences whose preferences they must anticipate. Yet current text-to-image exploration tools derive diversity entirely from the designer's own input–their prompts, their chosen dimensions, their search queries–confining exploration to what the designer already knows to look for. We present FocusGen, an interactive system that introduces external perspectives into visual design exploration through a "virtual focus group" of simulated persona agents. In contrast to prior persona systems in which multiple agents converge as critics on a single evolving artifact, FocusGen uses personas as parallel generators: each agent–constructed from demographic data, a procedurally generated backstory, and aesthetic preferences elicited through interviews–independently drives an iterative generation loop that produces its own visual concept, transforming one design brief into a spectrum of audience-conditioned directions. With real human participants, we confirm that the iterative refinement loop produces outputs people prefer over zero-shot generation. With synthetic agents at scale, we show that persona conditioning yields higher visual diversity than a generic-assistant baseline–measured by CLIP distance and corroborated by human perceptual judgments–and that open-ended preference interviews yield more diverse outputs than structured ones for both human and synthetic cohorts, while also revealing that agent cohorts recover only part of the diversity of comparable human cohorts. A qualitative study with 16 creative professionals suggests FocusGen helps designers discover unanticipated directions, overcome fixation, and probe audience contexts–while surfacing stereotyping risks that we analyze. We position FocusGen as a divergence scaffold for early-stage ideation rather than a substitute for audience research.
Read the full abstract

All authors Jaewon Choi, Helena Vasconcelos, Hyun Lee, Carolyn Zou, Tak Yeon Lee, Michael Bernstein

High Energy Astrophysical Phenomena

Hyperon bulk viscosity effects in neutron-star inspirals

The paper revisits the role of hyperon bulk viscosity during neutron-star inspiral. We find that hyperon bulk viscosity exerts only a minor influence on the gravitational-wave phase, causing a phase shift of $\sim 10^{-3}\,\rm rad$. This shift is too small to be detected by existing gravitational-wave observatories, and it seems unlikely that next-generation detectors will be able to isolate this effect from other factors. However, our analysis indicates that hyperon bulk viscosity can significantly heat the hyperon core of a neutron star, raising its temperature to approximately $5 \times 10^8\,\rm K$.
Read the full abstract

Artificial Intelligence

PhenoIntel: A Lifecycle-Aligned Multi-Agent Web Application for Verified, Accessible Plant Phenotype Analysis

Existing conversational plant-phenotyping platforms are difficult for plant scientists to use and lack the reliability scientific research demands: failed analyses are reported as valid measurements rather than flagged as missing, statistical tests run without checking assumptions, predictions carry no uncertainty estimate, and specialised hardware limits accessibility. We present PhenoIntel, a lifecycle-aligned multi-agent web platform that turns the full machine-learning workflow into a reliable, user-friendly phenotyping system. Nine specialised agents divide the analysis into stages, from image collection through model selection, inference, and reporting, rather than handing the whole task to one AI manager. Independent checks separate these stages, and every agent reads from and writes to one shared, fixed-structure record, so an inconsistent output from one stage is caught before it reaches the next. Uncertainty is matched to each model family, conformal prediction, detection-confidence spread, or Monte Carlo Dropout, rather than applied uniformly, and quality thresholds adapt to crop and task instead of one global cutoff. When no suitable model exists, PhenoIntel can propose, validate, and integrate a new one on its own. The model repository spans ten trained models across five crops and four imaging modalities. Classification models reach Macro F1 of 0.78-0.996; object-detection models reach 0.96 mAP@50 with a 54% reduction in counting error over an unoptimised baseline; and a temporal model reaches held-out Macro F1 of 0.7050. PhenoIntel runs in a browser on standard hardware, requiring no GPU, and a 1,200-test automated suite confirms complete pipeline execution. Every result carries calibrated uncertainty, validated statistics, and FAIR-compliant provenance, a combination existing conversational phenotyping tools do not offer.
Read the full abstract