Peer Revenue Equity in science

Papers

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

All of arXiv

Sunday 30 August

Statistical Methodology

Robust K-means Clustering using the Density Power Divergence Measure

We introduce a robust clustering method, MK-means DPD, that estimates cluster centers and covariance matrices using density power divergence (DPD) measures combined with Mahalanobis distance, making it resistant to outliers and adaptable to heterogeneous, elliptical clusters, unlike the classical K-means algorithm. Since Mahalanobis distance-based K-means lacks a general convergence guarantee, we further introduce a convergent variant, Density-Consistent MK-means DPD (DC-MK-means DPD), which redefines the cluster assignment step in terms of a pointwise DPD loss. We prove a formal theorem establishing that the resulting algorithm converges in a finite number of steps. We also propose two new robust internal evaluation indices, a Median Davies-Bouldin Index and a Trimmed Calinski-Harabasz Index, to ensure that performance comparisons are not themselves distorted by outliers. The efficacy of the proposed methods is demonstrated on simulated data, showing superiority over existing methods, and on two real datasets: Iris data, to identify similar species, and COVID-19 case fatality rate and infection rate data for countries worldwide, examining the resulting clusters' geographic and socio-economic patterns.
Read the full abstract

Computation and Language

Arkios: An Open Bilingual English-Nepali Language Model Trained From Scratch, with a Devanagari-Aware Tokenizer

We present Arkios, a 1.04B-parameter dense transformer pretrained from scratch on 150B tokens of bilingual English-Nepali text, using a custom single-file C/CUDA training stack and a Devanagari-aware byte-level BPE tokenizer built for this project. On ARC-Easy and ARC-Challenge, Arkios exceeds three comparably sized open models (Pythia-1.4B, TinyLlama-1.1B, OLMo-1B) despite an order of magnitude fewer training tokens, likely aided by a match between our educational-web-text pretraining data and ARC's grade-school-science format rather than a general capability advantage. We report full evaluation results under standard protocols, including a correction to an earlier partial-sample estimate, and findings specific to evaluating small models in a low-resource language: the standard multiple-choice-letter prompt format used by common evaluation harnesses places this model at chance on Nepali reading comprehension, and simultaneously at chance on English in the same format, which would lead a naive benchmark run to conclude the model has no Nepali ability when in fact it does. Concretely, both languages score at chance in the letter-choice format (0.240 Nepali, 0.236 English, against a chance baseline of 0.250), while scoring the answer text directly reveals genuine, English-favoring comprehension (0.306 Nepali, 0.387 English). We describe a manifest-conditioned tool-use contract introduced during instruction tuning, where tool calls are permitted only when a tool manifest is declared in context and suppressed otherwise, and report where that contract holds and where it does not. We release both the base and instruction-tuned model weights under Apache-2.0. The training code and a small privately-sourced portion of the Nepali pretraining corpus are not released; everything needed to reproduce the reported numbers from the released weights is included here.
Read the full abstract

Artificial Intelligence

VERA: Authority-Preserving Edge Revocation for Federated AI-Agent Workflows

Modern agent frameworks compose planners, tool agents, remote services, and shared specialists into runtime delegation graphs, but their revocation APIs still resemble token or subtree invalidation. When one delegation is withdrawn, the runtime must know which agents lose authority while independently authorized agents keep working. We study this authority consistency problem and introduce VERA (Verifiable Edge Revocation for Agents), a verifier-checkable revocation contract and API emitted by agent-runtime adapters as signed evidence. Under disjunctive authority, revoking edge e invalidates exactly T_intent(e,G) = reach(G) \ reach(G \ {e}), the agents whose every authorizing root path used e. Used as a contract, this target exposes two runtime failures: tree cascades over-revoke shared agents, while deployer-scoped cascades under-revoke cross-domain descendants. In a LangGraph framework-replt cells repeated 20 times yield 500compiled-framework traces and 2,000 valid signed delegation decisions; 13/25 cells contain runtime multi-parsharing and 8/25 contain cross-deployer shies 500/500 target proofs, preserves all320 alternate-parent shared-agent cases that tree cascade revokes, and rejects unauthorized signers and omission attacks. Baseline replay over 1,9that holder/node and tree-style targetscannot express this behavior. We further validate schema portability on A2A, AutoGen, and CrewAI artifacts: nine traces, including five executable Cregned delegation events that pass schema and signature checks.
Read the full abstract

Representation Theory

Transfer of Wakamatsu tilting modules along Frobenius extensions

Let $\iota: R\to A$ be a Frobenius extension and let $T$ be a Wakamatsu tilting left $R$-module. We give sufficient conditions for the induced module $A\otimes_R T$ to remain Wakamatsu tilting and establish an ascent–descent result for split centrally projective Frobenius extensions. We also characterize the tilting case by the vanishing of $\mathrm{Ext}_R^i(T,A\otimes_R T)$ for all $i>0$. Moreover, if $A\otimes_R T\in\mathrm{add}_R(T)$, then the natural map $S=\mathrm{End}_R(T)\to B=\mathrm{End}_A(A\otimes_R T)$ of endomorphism rings is a Frobenius extension and $T\otimes_S B\cong A\otimes_R T$ as $R$-$B$-bimodules. Applications to Brenner–Butler–Miyashita equivalences and some specific classes of Frobenius extensions are also discussed.
Read the full abstract

Classical Analysis and ODEs

On the Coefficients of Hurwitz-Type Matrix Polynomials

Consider the matrix polynomial $\mathbf f_n(z)=I_qz^n+A_1z^{\,n-1}+\cdots+A_n,$ where $A_j\in\mathbb C^{q\times q}$. Write $\mathbf f_n(z)=\mathbf h_n(z^2)+z\,\mathbf g_n(z^2).$ A matrix polynomial $\mathbf f_n$ is called a Hurwitz-type matrix polynomial if, for $n=2m$, $\mathbf g_n(z)\mathbf h_n(z)^{-1}$, and for $n=2m+1$, $\mathbf h_n(z)\bigl(z\mathbf g_n(z)\bigr)^{-1}$, admit finite continued fraction expansions with positive definite coefficients. We derive explicit formulas for the coefficients of Hurwitz-type matrix polynomials in terms of orthogonal matrix polynomials, Markov parameters, and Schur complements. We introduce the associated block Hurwitz matrix and establish determinant identities relating it to the corresponding block Hankel matrices. Finally, we settle a conjecture concerning the positivity of the determinants of the coefficient matrices of Hurwitz-type matrix polynomials. We prove that this property holds for degrees at most three, but fails in degree four by means of an explicit counterexample.
Read the full abstract

Machine Learning

A Lightweight Phenology-Aware YOLOv5 Framework for Tomato Growth Stage Detection in Resource-Constrained Bhutanese Greenhouse Environments

Accurate detection of tomato growth stages is essential for stage-specific greenhouse management and precision agriculture. In Bhutan, greenhouse cultivation is affected by altitude variability, large diurnal temperature fluctuations, diffuse illumination, limited automation, and a scarcity of locally annotated datasets, limiting the applicability of conventional deep learning models. This work proposes Pheno-Lite + Efficient Channel Attention (ECA), a lightweight, phenology-aware object detection architecture derived from Ultralytics YOLOv5 for tomato growth stage recognition. A balanced dataset of 2,464 annotated images was constructed from locally collected greenhouse images in Bhutan and publicly available tomato images, with augmentation designed to simulate local greenhouse conditions. The dataset includes vegetative (820), flowering (824), fruiting (820), and background (26) samples. The proposed architecture introduces two customized backbone modules: C3 PhenoLite, which enhances spatial and texture feature extraction using depthwise residual refinement, and C3 ECA, which strengthens inter-channel feature interactions through efficient channel attention. The proposed model achieves 90.6% precision, 88.8% recall, and 92.6% mAP@50, with 4.0 million parameters and 10.9 GFLOPs at 640 x 640 resolution. These results demonstrate its potential for real-time and climate-resilient greenhouse deployment in Bhutan.
Read the full abstract

Computer Vision

The Nearest Target Is the Wrong One: Target Separation in Arc2Face Identity Unlearning

Unlearning an identity from a face-conditioned generator by redirecting its conditioning embedding can silently fail if the redirected output is still verified as the original person. We show that this failure depends on a controllable choice of how far the redirection target lies from the forget identity in recognition space, and that the most intuitive target, the nearest neighbour, is the one most likely to cause it. We audit Arc2Face with a locked ArcFace protocol and a projection adapter that redirects identity conditioning before generation. On a hard-neighbour stress test built from the hardest 0.5% of eligible identities, four target-selection policies show a monotonic response: clean forgetting rises from 9/30 groups under the nearest hard target to 30/30 under the least similar one. Mean forget-identity re-identification falls from 51.9 to 0.0 while mean retention stays flat. This reflects successful redirection rather than outputs becoming unverifiable: 710 of 720 least-sim-hard generations arrive at the chosen target, with no leakage to unrelated identities. Re-verifying identical images with an independent recogniser (AdaFace) preserves that trend, correlating at r=0.94, arguing against a verifier artefact. Target separation is thus a first-order, reportable design variable for identity unlearning.
Read the full abstract

Computation and Language

When Does a Classifier Help an LLM? Classifier-Guided Prompting and Hybrid Classifier-LLM Models for Credit-Default Prediction

Credit-default prediction is an important task in financial decision making. Traditional methods use fitted classifiers such as logistic regression and random forests on tabular features. Large language models (LLMs) have recently been applied to this task through prompting. In this work we study how a fitted classifier and an LLM can be combined for credit-default prediction. We distinguish telling the LLM to imitate a classifier from using the classifier to build the prompt. We hypothesize that a fitted classifier can supply the ranking ability that an LLM prompt lacks. We experiment on the Default of Credit Card Clients dataset, and report recall, F1, and the area under the ROC and precision-recall curves, with bootstrap confidence intervals. We observe that a few-shot LLM has the highest recall (0.47) and F1 (0.50) of any single model but ranks worse than a random forest (AUC-ROC 0.72 against 0.79). Instructing the LLM to imitate a classifier gives no significant change. Pruning the prompt to the classifier's eight most important features raises recall by 0.071 and F1 by 0.032. Adding the classifier's predicted probability to the prompt raises the LLM's AUC-ROC from 0.72 to 0.78, matching the random forest, while keeping 0.118 higher recall than it. The reverse composition, and the use of several classifiers, do not help. We thus recommend a simple classifier-guided prompt for LLM-based credit prediction.
Read the full abstract

Populations and Evolution

Observation delays can bias inference of selective advantage in evolutionary competition

Relative-frequency trajectories are often used to infer selective advantage in competing biological populations. A common empirical approach is to fit a linear function to the logit-transformed frequency of an invading type and interpret the slope as the relative advantage. Here we test how this estimator is affected when the competing types are observed after type-specific delays. We use SARS-CoV-2 variant replacement in the United Kingdom as empirical motivation and study the mechanism with simple two-type models. In an ideal exponential replacement model, fixed or randomly distributed observation delays change the intercept of the observed log-odds trajectory but not its slope, provided that delayed counts are aggregated before frequencies are formed and boundary effects are absent. In nonlinear SIR-type models, where the relative growth is time-dependent, delays can compare the competing variants at different dynamical phases and substantially bias the fitted logit slope. This bias can occur even when the delayed replacement trajectory remains nearly linear on the logit scale. Thus, a good logistic fit to the observed data is not sufficient to guarantee that the inferred logit slope approximates well the relative growth advantage. Observation delays should therefore be accounted for when logit slopes are used as proxies for relative fitness in nonlinear evolutionary competition.
Read the full abstract

Computers and Society

AMINA: The Inclusive and Accountable AI for Marginalized Immigrant Nonprofit Assistance

Immigrant-led nonprofit groups, particularly those operating in politically sensitive contexts, face exclusion from formal registries and digital platforms. This paper reports a three-phase mixed-methods study with Iranian immigrant nonprofit practitioners: 27 semi-structured interviews, a co-design session, and 7 evaluation and feedback interviews on a prototyped AI assistant, AMINA. Our findings highlight how legitimacy barriers, capacity gaps, and politically charged misinformation constrain nonprofit operations. We translate these insights into design goals for an inclusive nonprofit AI assistant: support for everyday group operations, recognition of informal nonprofit efforts, proactive countering of misinformation, and multilingual, accessible interaction. User evaluations show AMINAs potential to reduce reporting burdens and foster transparency through proactive reminders, and catalyze collaboration across dispersed networks. We contribute to CSCW and HCI by characterizing the cooperative work of transnational immigrant nonprofits, extending scholarship on informality and misinformation, and demonstrating how AI can act as a collaborative partner that strengthens, rather than displaces, the human connections at the core of nonprofit ecosystems, while also posing major risks.
Read the full abstract

Cryptography and Security

Zero-Knowledge Predicate Proofs Between AI Agents: A Measured, Cross-Protocol Gateway and the Source-Integrity Gap

Multi-agent AI platforms move quickly from staging to production, but the way agents establish trust remains rudimentary: an agent either transmits raw data to a peer or accepts that peer's natural-language self-report that a value complies with policy. The first over-shares; the second is unverifiable and is exactly the channel prompt injection attacks. Prevailing responses emphasise identity, visibility, and post-hoc detection, and recent proposals for cryptographically enforced agent policy have been evaluated in simulation rather than execution. We take provable data minimisation between agents from proposal to running system. In our Zero-Knowledge Proof Gateway, agents exchange proofs of governance-defined predicates over private data rather than the data itself, so exposure is prevented by design rather than detected afterwards; because no interoperability protocol can carry such a proof, we propose a slot and implement it on both MCP and Agent2Agent from one endpoint. A 32-bit threshold predicate proves in 6.2 ms and verifies in 1.0 ms with a 608-byte Bulletproofs proof on one commodity vCPU; eleven adversarial experiments and nineteen protocol checks pass; and the system is deployed to Kubernetes with empirically verified network isolation. Our case study proves a retail client order is within its limit without revealing the amount, instantiating the GDPR data-minimisation principle as an enforced technical measure of the kind EU law now names explicitly. We then address the limitation no comparable work resolves: a predicate proof binds a statement to a committed value, never to the system of record. We give a construction fusing an enclave attestation with the proof in both directions, so verifying one artifact certifies jointly that the predicate holds and that the value was read by a specific measured binary, and test it against a mock authority.
Read the full abstract

Combinatorics

Mutually orthogonal anti-Latin squares

Anti-Latin squares were introduced in connection with non-linear secure network coding, and the extremal problem for large mutually orthogonal families is motivated by that setting. We study the maximum size $N_A(d)$ of a family of mutually orthogonal anti-Latin squares of order $d$. We prove that $N_L(d)+1\le N_A(d)\le N_L(d)+2$ for every $d\ge 3$, where $N_L(d)$ denotes the classical maximum size of a family of mutually orthogonal Latin squares of order $d$, and we show that in fact $N_A(3)=N_L(3)+1$ whereas $N_A(d)=N_L(d)+2$ for every $d\ge 4$. The upper bound is obtained by passing through balanced matrices, while the lower bound is given by a deterministic permutation argument. For all $d\ge 8$, and also for the exceptional order $d=6$, the upper bound is shown to be attainable by a general probabilistic construction. On the structural side, we show that a saturated family of size $d+1$ induces an affine plane of order $d$, and that the saturated case is characterized by the existence of an anti-coordinate grid decomposition; after transporting this condition to the fixed cell set $[d]^2$, it becomes a direction-completeness condition on the corresponding row-blocks and column-blocks. The remaining small orders are treated separately: $d=3$ is handled by direct analysis and classification of orthogonal triples, $d=4$ by an explicit saturated construction and an analysis of its finite-geometric structure, and $d=5$ and $d=7$ by explicit saturated examples arising from the random-grid framework. Thus $N_A(d)$ is determined in terms of $N_L(d)$ for every $d\ge3$, and its numerical value is obtained explicitly for every $3\le d\le9$.
Read the full abstract

All authors Eishiro Aoyama, So Hasegawa, Masahito Hayashi, Tomoki Sagara

Machine Learning

Tracing Generated Samples to Training-Data Clusters in Flow-Matching Models

Understanding which training samples influence a generated image is an important problem in generative modeling. In flow matching, training samples influence the generated image through the velocity field along the generation trajectory. Removing samples to examine their counterfactual influence changes the velocity field, and the resulting effect on the final image depends on how the change propagates through the trajectory. Consequently, local changes in the velocity field do not necessarily predict the final counterfactual effect. This work investigates attribution in flow-matching models through a hybrid analytical–learned approach, and uses it to derive trajectory-based attribution scores at the cluster level. We evaluate these attribution scores using independently retrained leave-one-cluster-out (LOCO) models, and compare with several attribution baselines using two different flow-matching latent spaces. Our experiments show that semantic similarity constitutes a strong baseline, while the closed-form trajectory-based attribution is competitive in some metrics without requiring counterfactual retraining or model gradients. Our results show that attribution in flow matching depends not only on semantic similarity to training samples, but also on the latent representation, trajectory dynamics, and how influence is propagated to the final output.
Read the full abstract

All authors Rania Briq, Ohad Fried, Michael Kamp, Stefan Kesselheim

Instrumentation and Methods for Astrophysics

CRUX: A topology-aware load balancer for mesh-based fluid dynamics codes on GPU clusters

The rapid growth of computational power has revolutionized numerical simulations, profoundly enhancing our understanding of fluids and plasmas. Computational fluid dynamics (CFD) simulations, which solve partial differential equations governing fluid or plasma motion on discretized grids, have been central to this progress. Recent advances have pushed the resolution and runtime of legacy numerical models to unprecedented levels while enabling newer codes to incorporate increasingly sophisticated physics. However, further scaling of these simulations has become a significant challenge, largely due to the comparatively modest improvements in networking capabilities relative to the rapid growth of floating-point performance in modern GPU-accelerated clusters. In this article, we introduce a novel load-balancing routine CRUX designed to scale efficiently for the most demanding CFD grids in astrophysics. Unlike traditional approaches based on space-filling curves, our method dynamically accounts for computational cost disparities among mesh blocks evolved with different timesteps while minimizing communication overhead. It is also able to take into account heterogeneous hardware. Through an extensive suite of benchmarks featuring up to 5,400 GPUs on OLCF Frontier and ALCF Aurora, we demonstrate that our load-balancing algorithm outperforms space-filling curve methods across all key metrics, including load uniformity, memory consumption, and communication efficiency, making it a robust solution for next-generation CFD simulations.
Read the full abstract

Cosmology and Nongalactic Astrophysics

Host Dependence and Line-of-Sight Effects on Galaxy-Galaxy Strong Lensing in Clusters

The cross section for galaxy-galaxy strong lensing (GGSL) events in galaxy clusters has repeatedly been found to be higher in observations than in cosmological simulations. We revisit this discrepancy using updated simulation methodology and investigate the dependence of the GGSL probability, $P_{\rm GGSL}$, on host-cluster lensing properties, baryonic physics, and correlated and uncorrelated line-of-sight structure. We find that correlated material within $\sim 35$ cMpc of the cluster along the line of sight enhances $P_{\rm GGSL}$ by a few percent for typical systems and by up to $\sim15\%$ for the most efficient lenses. At fixed cluster mass, dark-matter-only simulations yield GGSL probabilities up to an order of magnitude lower than hydrodynamical simulations. We also find a strong dependence on the host-cluster Einstein radius, with an approximate scaling $P_{\rm GGSL} \propto \theta_{\rm E}^{2}$. Matching simulated and observed clusters in both mass and Einstein radius seems to reduce the discrepancy relative to previous comparisons. However, our analysis does not clearly resolve the GGSL discrepancy, as the inferred tension depends strongly on the field-of-view definition: square fields are approximately consistent with simulations, while cluster member-bounded fields yield observed probabilities a factor of $\sim 2-3$ higher. Until observations and simulations share a matched selection function and matched measurement methodology, the residual tension cannot be cleanly attributed to either astrophysics or cosmology.
Read the full abstract

All authors Cian Roche, Michael McDonald, Priyamvada Natarajan, Keren Sharon, Isaque Dutra, Massimo Meneghetti, Mark Vogelsberger

Number Theory

Sets whose differences avoid a bracket quadratic

Suppose a set of integers $A\subseteq\{1,\dots,N\}$ has no solutions to $a-a'=n\lfloor\sqrt[3]2n\rfloor,$ for distinct $a,a'\in A,$ and $n\in \mathbb{N}.$ We show that $|A|\ll N^{1-c}$ for some absolute constant $c>0.$ To do this, we prove quantitative bounds on the van der Corput property for certain sets of bracket quadratics. This comes as a consequence of establishing exponential sum estimates for these sets, utilising a theorem of Green and Tao on the quantitative equidistribution of polynomial orbits on nilmanifolds, closely following the approach of Neale who went on to prove a Waring-type result. We also extend our result to differences avoiding a family of bracket polynomials (also known as generalised polynomials).
Read the full abstract

Differential Geometry

Lie groupoid integration of singular isometries of the Poincaré disk

For every $n \geq 1$ there is a distinguished $\mathfrak{sl}_2(\mathbb{R})$ action on the Poincaré disk, arising as the infinitesimal isometries of a hyperbolic metric with conical singularity of order $n-1$ at the origin. For $n=1$ this is the standard infinitesimal Möbius action, and for $n>1$ these vector fields have singularities at the origin and are incomplete, preventing integration to a global Lie group action. However, they naturally define an action Lie algebroid $\mathcal{A}_n=\mathfrak{sl}_2(\mathbb{R})\ltimes \Dbarstar $ over the punctured disk. We construct an explicit Lie groupoid $\mathcal{G}_n$ integrating $\mathcal{A}_n$ and compare it to the \v Severa–Weinstein groupoid. Although $\mathcal{G}_n$ is not an action groupoid, its restriction to the boundary recovers an $n$-fold Möbius action on the boundary circle.
Read the full abstract

Computation and Language

Budget-Aware Compression Pipeline for Single-GPU LLM Inference: Methods, Trade-offs, and Coupling Effects

Single-GPU deployment of 70B-parameter language models on an NVIDIA GPU is constrained by device memory, long-context throughput, and engineering integration cost. We cast single-GPU inference as a budget-aware design problem over these three axes and study how pruning, quantization, and KV-cache compression interact under realistic execution. Controlled ablations show that layer-wise pruning makes weight quantization more robust. KV-cache sparsification complements INT8 KV quantization by reducing memory without hurting decoding speed, while static vector quantizers often conflict with dynamic caching. Guided by these coupling results and explicit budget tracking, we assembled a practical pipeline and compressed a 70B model to about 33 GB, sustained about 57 tokens/s on 10k token prompts on a single A40, and kept absolute accuracy within 5% on common and reasoning benchmarks. We contribute design rules and a reproducible evaluation protocol that jointly report quality, memory, and end-to-end speed, and we provide a foundation for automated pipeline search under realistic single-GPU constraints.
Read the full abstract

Quantum Physics

Newtonian Gravitational Curvature-Induced Entanglement Generation

We show that gravitational curvature can control the generation of nonlocal quantum correlations in a hybrid qubit-mechanical device. The tidal field of a nearby source mass modifies the susceptibility of a shared mechanical oscillator, thereby tuning an oscillator-mediated qubit-qubit interaction and the resulting entangling phase. An exact treatment of the dynamics reveals stroboscopic geometric gates whose accumulated phase is directly sensitive to gravitational curvature. Treating the curvature as an unknown parameter, we derive the ultimate quantum limit for its estimation and identify a parity-based measurement that saturates this bound. Solving the full master equation with the mechanical mode retained explicitly, we find that thermal occupation of the mediator suppresses entanglement between the closure times but is undone at each closure, exactly in the unitary limit for any initial mechanical temperature, so that ground-state cooling of the oscillator is not a prerequisite for the protocol. Mechanical damping and qubit dephasing behave differently: they leak branch information irreversibly to the environment, and it is the heating and dephasing rates, rather than the bath occupation alone, that limit the entanglement visibility and the number of usable interrogation loops. In contrast to gravity-mediated entanglement proposals, entanglement is generated by the mechanical oscillator while gravity acts solely as a classical control field. We quantify the achievable curvature parameters; all curvature dependencies are analytic, allowing for exact rescaling of the results. The scheme therefore demonstrates gravitational control of a quantum interaction and provides a route to curvature sensing based on a nonlocal two-qubit phase rather than on local phase measurements.
Read the full abstract

All authors Mughees Ahmad Khan, Asad Ali, M.I Hussain, H.Kuniyil, Saif Al-Kuwari

Statistics Theory

Distribution-free testing of linear type

We introduce a distribution-free goodness-of-fit test, termed the omega-1 test, which naturally complements the Kolmogorov–Smirnov test and Cramér–von Mises test and can be viewed as their (piecewise) linear analog. Defined as an $\mathrm{L}^{1}$-functional of the empirical process, the test statistic improves on balancing sensitivity to localized and diffuse alternatives and gives a robust and interpretable measure of distributional discrepancy, apart from close connections to the Wasserstein 1-distance. For finite samples, we derive a finite-dimensional computational form for the statistic under general conditions, which leads to various explicit formulas for its null distribution. Under mild continuity assumptions, the limiting statistic is distribution-free, with explicit distribution formulas. In composite settings, the statistic is also compatible with the Khmaladze transformation, enabling asymptotically distribution-free testing. The limiting transformed statistic also has an explicit distribution that escapes reliance on intractable compensator processes or purely numerical evaluation. Simulation results indicate rapid convergence of the finite-sample distributions to their limiting counterparts and support the practical applicability of the test.
Read the full abstract

Machine Learning (Statistics)

Learning Representations through Token Prediction: Geometry, Approximation, and Downstream Guarantees

Token prediction is a central pre-training objective for modern language models. Despite its empirical success, why token prediction learns broadly useful representations remains incompletely understood. We develop a statistical framework connecting token prediction with representation geometry, encoder approximation, and downstream performance. Under a softmax prediction head, we show that accurate token prediction organizes token embeddings according to similarities between the distributions of contexts in which different token types appear, as measured by Hellinger distance, with explicit errors governed by prediction accuracy and token frequency. Meanwhile, the contextual representation provides a low-dimensional coordinate for the conditional distribution of the target token relative to these embeddings. We further introduce a self-consistency principle showing that repeated applications of a shared representation block can progressively refine the contextual representation without introducing additional block parameters. Among representations with the same prediction accuracy, this recurrent construction favors those that can be stably reconstructed from their contexts. Finally, we establish downstream guarantees for token generation, token community recovery, and classification by a linear probe, showing how prediction accuracy and recovered geometry translate into performance beyond the pre-training objective. Together, these results explain how the simple objective of predicting tokens can recover semantic geometry and produce broadly useful representations. A controlled simulation illustrates the theoretical mechanisms.
Read the full abstract

Spectral Theory

Uniform Non-Localization under Robin Boundary Perturbations: Spectral Splitting and Bounded Eigenspace Complexity

We study the high-energy spectral effect of Robin boundary perturbations on complete Laplace eigenspaces and its consequences for eigenfunction non-localization. Highly degenerate Neumann levels generally split under Robin perturbation, but this splitting need not produce a simple spectrum: distinct modal classes may still contribute to the same Robin eigenvalue. We prove that, for every fixed positive Robin parameter, the number of modal classes contributing to any one eigenvalue is uniformly bounded over the spectrum on the equilateral triangle and on every rectangle with rational squared aspect ratio. On the equilateral triangle, for all sufficiently small Robin parameters, the Neumann complete-eigenspace observation estimate persists with a constant uniform both in the eigenvalue and in the boundary parameter. For an arbitrary fixed positive Robin parameter, a high-frequency shell-splitting analysis gives a spectrum-wide bound on the number of modal classes contributing to any one Robin eigenvalue. In the nonsquare rectangular case, the corresponding splitting is governed by a strongly convex profile on weighted quadratic shells. Since each modal class has uniformly boundedplane-wave complexity, the spectral complexity bound implies observation on every measurable set of positive measure through a multidimensional Turán–Nazarov inequality, without a frequency-separation assumption. Thus, spectral simplicity is not required for complete-eigenspace non-localization: uniformly bounded spectral coincidence complexity is sufficient.
Read the full abstract

Machine Learning

Selection, Representation, and Execution in Sparse Fourier Neural Operators

Sparse representations are often expected to make models smaller and also reduce inference cost. For Fourier Neural Operators (FNOs), these objectives are not equivalent or do not always align: removing parts of the learned operator can leave the underlying transforms and dense computations unchanged, while changing the grid on which the model is evaluated can introduce overhead of its own. We therefore distinguish sparsity in the representation, in the stored parameters, in the theoretical operation count, and in measured runtime, and present an empirical study of several routes toward sparse FNOs that tests each transition between them separately. Coarsening the execution grid reduces the theoretical cost without reducing measured latency, and adding a correction term recovers accuracy at the cost of making the model slower. Even an 83% parameter reduction remains slower than the dense baseline under ordinary execution. These results motivate a stricter definition of useful sparsity: the deployed operator must preserve solution accuracy and map its reduced support to a genuinely cheaper execution path.
Read the full abstract

Number Theory

Matrix representations and arithmetic properties of jacobsthal numbers via binary 3x3 matrices

We study matrix representations of the Jacobsthal sequence generated by binary 3x3 matrices with determinants 0, 2, and -2, using linear algebraic methods analogous to Fibonacci-type constructions. Explicit formulas for matrix powers are obtained, yielding identities for Jacobsthal numbers, including convolution formulas, trace relations, determinant expressions, and Cassini-type identities. We further derive congruence relations and recurrence formulas, and analyze arithmetic properties such as partial sums and the Sidon-type structure of the sequence. Finally, we prove that exactly three conjugacy classes of binary 3x3 matrices generate the Jacobsthal sequence, providing a unified algebraic framework that connects matrix theory with second-order linear recurrences.
Read the full abstract

Computer Vision

TAKE 85: Testing Audiovisual filmmaKer's intEnt across 85 Hours of Film

Films communicate through deliberate creative choices, including lighting, color, composition, editing, dialogue, music, and sound. Humans naturally interpret these signals as directorial intent, yet current multimodal large language models (MLLMs) are evaluated almost exclusively on understanding what happens rather than why it is presented that way. We introduce TAKE 85, the first benchmark for directorial-intent understanding, comprising 398 short films (85 hours) with expert-verified question-answer pairs spanning global and fine-grained visual and audio intent. Through controlled modality ablations, TAKE 85 enables systematic evaluation of multimodal reasoning. Experiments on state-of-the-art MLLMs reveal a substantial gap between perceptual recognition and intentional understanding: while models accurately describe events and narratives, they consistently fail to infer the communicative role of filmmaking decisions. Our results establish directorial intent as a previously overlooked dimension of multimodal understanding: even the strongest model reaches only 58 out of 100, and our ablations show that no input modality is sufficient on its own. All code, Q&As, and models are publicly available from https://github.com/KaiShinozakiConefrey/Take-85
Read the full abstract

All authors Kaishuu Shinozaki-Conefrey, Olivier Pascaud, Robin Courant, Xi Wang, Dimitris Samaras, Vicky Kalogeiton

Machine Learning

How do World Models and Policies Compose in LLM Agents? A Joint Spectral and Behavioral Account

How do LLM agents come to both understand environments they act in and master tasks set within them? Through controlled experiments combining world-model training (next-state prediction) and policy training (reward maximization), we investigate this question. We dissect the resulting models through their additive parameter updates. Geometrically, we find effective world-model updates are low-rank and share an input-feature subspace with policy updates while writing to nearly orthogonal output directions, whether trained separately or sequentially. However, we find that, in projection interventions, the sequential update induces more robustness than separate policy RL when removing the world model's leading input directions, suggesting that it has learned alternative input pathways. Behaviorally, we find the sequentially trained agent explores a wider range of states and actions. Based on this, we ask: does policy training preserve world knowledge as well as it could? We probe this with training-free merging built on the geometrically motivated input basis plus an online world-model loss during policy RL, and show both improve over the untreated baseline. Our findings suggest world knowledge and task-directed ability can be learned in geometrically complementary forms, and that future post-training pipelines should consider how best to engineer the interface between them.
Read the full abstract

All authors Ruize Xu, Xiao Yu, Yujin Tang, Chenming Shang, Nikhil Singh

Human-Computer Interaction

Occlusion-induced risk and interventions in pedestrian-autonomous truck interactions on multi-lane roads: A virtual reality study

Autonomous trucks (ATs) may introduce distinct pedestrian-safety risks because of their large physical dimensions, constrained braking capability, limited driver-based communication cues, and potential to occlude surrounding traffic. This study employed a controlled virtual reality experiment with 54 participants to investigate pedestrian-AT interaction risk in an unsignalized multi-lane crossing scenario and to evaluate occlusion-targeted risk mitigation strategies. The experiment examined the effects of near-side vehicle type, weather condition, and far-side vehicle yielding strategy on pedestrian behavior, perceived risk, and objective safety. Based on a representative high-risk scenario, three targeted interventions were designed and tested: an environment-aware external human-machine interface (eHMI), a projected eHMI, and an auditory warning. The results showed that ATs increased perceived risk and encouraged more cautious crossing behavior, suggesting a risk-compensation effect. However, this compensation was weakened under rainy conditions, where braking-related safety margins were reduced. AT-induced occlusion further increased far-side interaction risk by limiting pedestrians' recognition of hidden vehicles. Among the three interventions, the projected eHMI showed the best overall performance, improving objective safety margins, enhancing risk awareness, and supporting behavioral adjustment. These findings highlight the need for AT-specific interface and warning strategies that address both intention communication and risk localization.
Read the full abstract

All authors Yun Ye, Yuan Che, S.C. Wong, Stergios-Aristoteles Mitoulis, Haoyang Liang

Computation and Language

Pak3H: Evaluating the Cost of Cultural Mismatch in LLM Alignment with a Human-Contextualized Urdu Benchmark

Large language models (LLMs) demonstrate strong Helpfulness, Harmlessness, and Honesty (3H) alignment in English-centric settings, but these gains transfer poorly to low-resource languages due to cultural mismatches. Existing multilingual 3H benchmarks rely predominantly on automated translation or LLM based synthesis, propagating source-language biases while sacrificing local relevance. To address this gap, we introduce Pak3H1, the first human-validated, culturally contextualized Urdu benchmark suite for 3H alignment, comprising PakAlpaca (helpfulness), PakBeaverTails (harmlessness), and PakTruthfulQA (honesty). Our multi-stage pipeline integrates manual cultural adaptation and dictionary-guided post editing to prioritize native speaker judgment, ensuring both semantic fidelity and contextual authenticity. Zero-shot evaluations across multiple open and proprietary LLM architectures reveal systematic cross-lingual alignment gaps: helpfulness win rates decline under localized contexts, harmlessness guardrails break down against regional safety risks, and composite honesty metrics degrade substantially due to localized factual constraints. These findings expose structural limitations in current alignment approaches, underscoring the necessity of human-guided localization for equitable multilingual evaluation.
Read the full abstract

Materials Science

Libron-phonon coupling and hydrogen-bond dynamics in the vacancy-ordered perovskite (NH4)2SnCl6: a temperature- and pressure-dependent Raman study

Vacancy-ordered perovskites R2MX6 combine a rigid inorganic framework with a molecular A-site cation, but how the cation dynamics couples to the lattice phonons under external perturbation remains largely unexplored. The librational motion of NH4+ in (NH4)2SnCl6 is strictly Raman-silent by symmetry, which is why it has been probed almost exclusively by neutron scattering, NMR and NQR. We show that it is nevertheless accessible to Raman spectroscopy, through the renormalization it imposes on the allowed [SnCl6]2- modes. Combining single-crystal X-ray diffraction with Raman scattering between 10 and 300 K and up to 10.1 GPa, we find that the average cubic structure varies smoothly with no anomaly, while below ~100 K the [SnCl6]2- modes acquire a libron-phonon renormalization (E_eff ~ 4.7 meV) and a symmetry-selective line asymmetry, and the N-H stretch passes through a minimum near 120 K (E_eff ~ 9.3 meV); both track the classical-to-quantum crossover of the ammonium rotor. The two effective energies lie below the bare librational transition of 13.4 meV measured by neutrons, as expected for a self-energy scale. The ammonium linewidths, by contrast, are governed by pure dephasing rather than by anharmonic decay and carry no crossover signature. Under pressure the octahedral modes stiffen smoothly, whereas the cavity responds twice: the translational F2g mode of NH4+ gains Raman intensity above 1.3 GPa, and the N-H stretch inverts its pressure slope near 1.7 GPa, with no change of space group and full recovery on decompression. The rigid octahedral framework is thus essentially decoupled from a dynamically active NH4+ subsystem that carries the response to both perturbations, offering a route to tune cation-phonon coupling independently of the octahedral network.
Read the full abstract

All authors Vasco S. Neto, Mayra A. P. Gómez, Bruno S. Araújo, Alejandro P. Ayala