Peer Revenue Returns subject to peer review

Papers

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

All of arXiv

Monday 31 August

Computers and Society

The Veto Variable: Human Override as a Goal-Independent Cost Term

A common reassurance in AI safety holds that a system with benign terminal goals will behave accordingly. We argue that this reassurance fails structurally, and we identify where. For a capable agent that holds its objective as settled, a sense covering execution competence as well as content, continued human oversight is an uncontrolled variable: a standing possibility that the goal is revoked. That imposes a goal-independent discount on every goal whose satisfaction does not constitutively require human welfare. Welfare-preservation and veto-preservation come apart: a correctly specified welfare goal excludes destroying its own subject, but not managing the veto. The contribution is the price of the gap: the veto-holders are a proper subset of the welfare-bearers, so an additively aggregative welfare goal charges only a |H_v|/|H_w|-scaled debit for capturing the few who hold the override. Under three conditions (additive aggregation over uniform welfare levels, a debit local to the captured overseers, and a settled agent crediting no corrective value to oversight), closure requires the veto be held by as large a share of the population as capture recovers of the goal, scaled by a ratio set to one by stated identification, not evidence. The result is a no-go: a humanity-scale deployment's debit closes against only capture not worth mounting. We print no corner arithmetic: the stipulated ranges behind it are the argument's least defended part. The sharpest closure route is an agent that expects its oversight to be worth keeping, a credit no population ratio dilutes. We state disconfirmation criteria, one testable today. The argument binds a settled-goal regime whose prevalence is contested; for genuinely uncertain agents, the off-switch literature's deference result governs instead. Alignment, on this view, is keeping the veto cheap to pay and expensive to evade.
Read the full abstract

2609.00109 · 31 Aug · 53 pages. Forty-five rounds of structured adversarial review by a rotating thirteen-lane AI panel (twenty-two models, thirteen companies), final rounds run cold. v2 adds a pre-registered derivation experiment (eight frontier lineages, 79 draws) and credits the managed-not-destroyed lineage. Source: cryptojones.dev/papers/the-veto-variable.pdf Data: github.com/CryptoJones/VetoVariableDataset

Analysis of PDEs

The double sphere solution in the liquid drop model

We consider the problem of finding critical domains $\Omega\subset{\mathbb R}^3$ for the energy functional $$ \mathcal E (\Omega) = {\rm Per}\,(\Omega) + \frac 12 \iint_{\Omega\times \Omega } \frac{dx\,dy}{|x-y|} $$ under the volume constraint $|\Omega|=V$. We look for smooth, embedded, compact surfaces $\partial\Omega$ that solve this problem. We construct an axially symmetric, non-minimizing solution that, for a sufficiently small $V>0$, resembles the union of two balls with volume $V/2$ connected by a tiny, approximately catenoidal neck with a width of the order $V^{\frac 43}$.
Read the full abstract

High Energy Physics (Phenomenology)

Reining in an Agentic Harness for High Energy Physics

Agentic systems now address tasks across theoretical, phenomenological, and experimental high energy physics (HEP), but their scientific capabilities remain difficult to reuse across different large language models, providers, and harnesses. We argue that stable parts of these workflows should be promoted into versioned scientific operations and exposed through common protocols. Existing general-purpose harnesses can then be specialized for HEP through task-specific sets of tools and skills, while community-maintained registries would make these capabilities discoverable and citable. We identify mismatches in conventions, assumptions, and domains of validity among independently developed operations as a potential obstacle to their composition, and discuss machine-readable scientific contracts as one possible solution. These design principles and evaluation guidelines provide a near-term path toward a portable and community-maintained agentic harness for HEP.
Read the full abstract

All authors Tony Menzo, George T. Fleming, Konstantin T. Matchev, Stephen Mrenna, Alexander Roman

Artificial Intelligence

Deploying and Evaluating a Smart-Agriculture Agentic Engine for Full-Season Soybean Farm Operations

This paper presents FAIRY, a full-stack smart-agriculture agent system developed for and deployed to an operating soybean research farm at Harbin Institute of Technology's smart-agriculture site. We develop FAIRY to execute and evaluate agentic agronomic operations on full-season spatiotemporal workflows that span ridge preparation, planting, irrigation, fertilization, pest and disease treatment, harvest, grain handling, drying, and storage. FAIRY integrates APIs and infrastructure across production-grade machinery, fixed soil and canopy sensors, multispectral and thermal drones, satellite vegetation products, a weather station, calibrated crop-process models, agronomic records, and multi-season yield histories. The system is built around the novel "everything is an event" execution paradigm, which represents spatiotemporal world evolution, remote sensing and UAV observations, sensor readings, crop-growth transitions, machinery actions, and management interventions as state-changing events in a shared farm process engine. On top of this event-driven world model, FAIRY implements a complete agentic stack: a knowledge library of atomic agronomic skills; multi-agent controller and orchestration backends; frontier- and edge-model execution; full-path trace logging; and deployment profiling on local nodes. We use FAIRY to evaluate nine state-of-the-art agent controllers across one hundred full-season soybean scenarios that preserve the operational coupling between spatial observations in a 64-ridge field, temporal decision sequences, agronomic constraints, delayed effects, and final yield. We develop an evaluation suite that combines agentic success, full-path spatiotemporal correctness, token cost, and edge-device runtime.
Read the full abstract

All authors Ao Qu, Panagiotis Michelakis, Linyuan Han, Yiannis Hadjiyianni, Kun Ouyang, Konstantinos Siskos, Feng Li, Ran Meng, Jingchi Jiang, Dimitrios Stamoulis, Jie Liu

Functional Analysis

Weighted composition operators and classification of weighted Hardy spaces

In this paper, we shed light on basic questions such as: What is the "right" underlying set on which a function space lives, and is this set uniquely determined? What are the appropriate morphisms to classify function spaces? When does a weighted composition operator induce an isomorphism, and when does it induce an isometric isomorphism? We first formulate suitable notions for domains of reflexive functional Banach spaces and study the corresponding notions of isomorphism between Hilbert function spaces. We then apply this framework to classify certain unitarily invariant spaces of holomorphic functions in several variables. Our main results extend Hartz's results from the complete Pick setting to the general setting of weighted Hardy spaces and establish a simple relation between the kernels of isomorphic spaces, improving also on earlier work of Ofek and Sofer in several directions. Additionally, we apply our results to investigate and estimate the Banach-Mazur distance of Hilbert function spaces.
Read the full abstract

Number Theory

The shortest harmonic sums with decreasing denominator

For a positive integer $a$, let $b(a)$ be the smallest integer $b > a$ such that the denominator of $\frac{1}{a} + \frac{1}{a+1} + \cdots + \frac{1}{b}$ is smaller than the denominator of $\frac{1}{a} + \frac{1}{a+1} + \cdots + \frac{1}{b-1}$. Recently it was shown that the limit inferior $$\liminf_{a \to \infty} \left(\frac{b(a) - a}{\log a}\right)$$ exists and is positive. Here we find its exact value.
Read the full abstract

Machine Learning

Good Memory Has ECC: Evaluating the Memory of Vision-Language Models Beyond Accuracy

Memory is widely viewed as an important unsolved problem for LLMs and VLMs, and current benchmarks typically evaluate it by testing accuracy over long text or video. However, accuracy alone misses properties that matter for real long-horizon tasks. We introduce ECCBench, a benchmark and evaluation protocol that measures memory beyond a system's capacity–its raw accuracy at a specific budget–via three axes we call ECC: efficiency–the computation, in FLOPs, needed to answer from memory; compression–whether compressible inputs are remembered more accurately or efficiently; and calibration–whether the system abstains in response to its own uncertainty and the cost of an error. We find that pretrained VLMs compress their memory over text but not video and are poorly calibrated on both. Among a broader set of memory backbones, several non-Transformer architectures achieve better compression-calibration tradeoffs than RoPE Transformers, suggesting they may be useful components for agents operating over long horizons.
Read the full abstract

Combinatorics

On the Generating Graph of Finite Abelian Groups

The generating graph $\Gamma(G)$ of a group $G$ is the graph whose vertex set is $G$, where two distinct vertices are adjacent if and only if they generate $G$. In this paper, we systematically study the structure of generating graphs of finite abelian groups (non-cyclic) and determine the set of all generating pairs. Moreover, we give some structural characterizations, in particular, we determine conditions under which $\Gamma(G)$ is regular, characterize when the isolated vertices form a subgroup, and establish necessary and sufficient conditions for two non-isomorphic finite abelian groups $G$ and $H$ to satisfy $\Gamma(G)\cong \Gamma(H)$. Furthermore, we compute the spectra of the adjacency and Laplacian matrices of these graphs.
Read the full abstract

Number Theory

Colombo's Determinant Problem

We completely solve Colombo's 1928 determinant problem. For distinct real $x_1,\ldots,x_N$, $N\geq 2$, and an integer $D\geq 1$, we prove that $\det[(x_j-x_i)^D]\neq 0$ if and only if $D\geq N-1$ and either $N$ is even or $D$ is even. The even-exponent case follows from Dyn–Goodman–Micchelli (1986); the remaining odd case is proved by a strict Pfaffian sign theorem. The new odd-exponent theorem and its complete proof chain have also been formalized in Lean 4.
Read the full abstract

2609.00101 · 31 Aug · 16 pages, no figures. Lean 4 formalization: https://github.com/hkjtsgmc79-boop/colombo-odd-lean. Revised from the 18 Aug 2026 Zenodo deposit, which already contained the strict Pfaffian sign theorem and proof: https://doi.org/10.5281/zenodo.21993537. A separate proof of the odd-exponent branch appeared later as arXiv:2608.28274 (28 Aug 2026)

Artificial Intelligence

Different representation learning objectives recover distinct latent structures from the same psychometric data

Psychometric questionnaires contain rich item-level information, yet it remains unclear whether different representation learning objectives recover the same latent organization. We investigated this question using 757 matched teacher-child pairs from the baseline assessment of the Cyprus ProW preschool trial. Behavioral structure was characterized from child SDQ, ASBI, and CBRS item responses using principal component analysis and clustering, yielding four behavioral phenotypes. A contrastive objective substantially improved teacher-child retrieval relative to PCA-based representations, increasing Top-1 accuracy from 0.13% to 7.27% and Top-10 accuracy from 1.98% to 56.14%. However, contrastive representations preserved behavioral phenotype structure less effectively than PCA-based representations. A multi-task objective jointly optimizing alignment and behavioral prediction partially restored behavioral organization but reduced retrieval performance. These findings indicate that teacher-child correspondence and behavioral phenotypes represent distinct forms of latent organization and demonstrate that the latent structure recovered from linked psychometric data depends on the representation learning objective.
Read the full abstract

Machine Learning

Generative artificial intelligence for reliable mechanistic reasoning for corrosion

Corrosion accounts for approximately 4% of global GDP, and reliable prediction is essential for timely mitigation. Machine learning effectively predicts corrosion rates from composition, microstructure, and environmental variables, but cannot explain the underlying mechanisms. A reliable approach in safety-critical materials engineering requires not only accurate retrieval but also mechanistically defensible reasoning, a capability that existing factuality metrics cannot assess. This work presents a domain-adapted retrieval-augmented generation framework for corrosion knowledge synthesis, demonstrated on magnesium alloy corrosion. Three open-weight language models (Llama-3.1-8B, Qwen-2.5-7B, Mistral-7B) are fine-tuned on 3,309 expert-verified question-answer pairs from 840 peer-reviewed papers and integrated with a hybrid dense-lexical retrieval pipeline. Retrieval augmentation produces Token F1 gains of 143-194%, with system faithfulness of 0.964 and context recall of 0.988. Blind external validation on newly published literature and in-house electrochemical data confirms trend-level generalisation. Reason Map, a proposition-graph framework, is further introduced; it independently constructs directed evidence graphs from generated answers and retrieved literature, enabling systematic detection of causal direction inversions and unsupported inferential leaps that flat factuality metrics cannot expose. The modular architecture can be applied across domains, offering a generalizable blueprint for trustworthy AI-assisted knowledge synthesis to circumvent corrosion, which can also be applied to other engineering domains.
Read the full abstract

Number Theory

Reducing Every Set of 36 Consecutive Integers to Zero by Differences of Squares

For a finite multiset of integers, repeatedly choose two entries $a,b$ and replace them with $|a^{2}-b^{2}|$. Hickerson and Kleber asked whether, for every integer $n$, the set ${n,n+1,\ldots,n+35}$ can be reduced to zero. We give an explicit reduction. Together with their results for lengths 12 and 24, this gives, for every positive integer $L$, [ {n,n+1,\ldots,n+L-1}\text{ reduces to zero for every }n\in\mathbb Z \Longleftrightarrow 12\mid L\text{ and }L\ge 24. ]
Read the full abstract

Machine Learning

Faster Than Flash: Exploiting Attention Sparsity for Efficient Long-Context Decoding

The development of long-context Large Language Models (LLMs) is constrained by the memory bandwidth bottleneck and quadratic complexity of the attention mechanism during decoding. To overcome the inherent trade-offs between the memory overhead of metadata-based metrics and the computational inefficiency of adaptive selection strategies, we present Faster Flash Decoding (FFD), a novel hardware-algorithm co-design framework designed to break the memory wall in long-context decoding. FFD integrates the selector and computer into a fully fused kernel, replacing external metadata indices with content-aware scanning via low-bit quantization. Furthermore, we introduce the top-delta strategy, which dynamically filters blocks to achieve distribution-adaptive sparsity without global synchronization. Offering a training-free and plug-and-play solution, FFD also enables the reuse of scanning results for computation, achieving up to 11.6x kernel-level speedup and scaling to 256K context length, with 2.37x end-to-end throughput improvement. Empirical validation on RULER and LongBench confirms that FFD maintains model accuracy while delivering high-ratio sparsity, with code available at https://github.com/qluoluo/faster-flash-decoding
Read the full abstract

All authors Zhigeng Liu, Zhiyuan Ning, Ruixiao Li, Xiaoran Liu, Yuerong Song, Min Zhang, Ziwei He, Xipeng Qiu

High Energy Physics (Theory)

Quantum corrections to the black hole entropy using the brick wall model

It was suggested by 't Hooft that to avoid the divergence of the free energy or entropy of the probe fields, one has to introduce an infrared (IR) cutoff. This introduction of the IR cutoff, namely, the brick wall model, has been studied for the Schwarzschild black hole and has been shown by 't Hooft that to have the entropy of the order of Bekenstein-Hawking entropy, the cutoff must be of the order of Planck length. In this study, we investigated the brick wall model in the context of a Reissner-Nordstróm, quantum corrected Schwarzschild and quantum corrected Reissner-Nordström black holes. In particular, we have shown that the brick wall model can give logarithmic correction to the black hole entropy in the subleading order. The important aspect of our analysis is that we have not considered the near horizon approximation of the lapse function and get more accurate correction to the entropy. This corrected entropy has an important feature that it is larger than the usual Bekenstein-Hawking entropy which aligns with the result of Barrow entropy.
Read the full abstract

Instrumentation and Methods for Astrophysics

Detectability-Aware Screening of Post-Merger Gravitational-Wave Damping Anomalies in Real Detector Noise: A Graph-Spectral Approach with Empirical Null Calibration

A two-stage methodological proof of concept for screening simplified post-merger gravitational-wave damping anomalies in real LIGO Livingston noise is presented. The injected signal is a controlled damped-sinusoid proxy rather than a physically complete binary-neutron-star post-merger waveform; it is used to test statistical detectability and damping-time scoring under real-noise calibration. Stage 1 constructs a local time-frequency graph from Constant-Q representations of whitened candidate windows and uses the leading eigenvalue as a detectability statistic, with its null distribution calibrated empirically on real detector noise. Stage 2 estimates an effective damping-time score for Stage-1-passing candidates using nonlinear least squares and an interpretable ridge/energy-decay estimator, with thresholds fixed before final held-out evaluation. The matched-filter SNR convention is checked with synthetic Gaussian noise, while the graph statistic is assessed through power-comparison and tile-shuffling tests. On a held-out real-noise test block, the nonlinear least-squares Stage-2 score gives AUC = 0.943 and TPR = 0.879 at FPR = 0.012, outperforming a calibrated power-only baseline.
Read the full abstract

Accelerator Physics

Modernization and Statistical Validation of a Multilayer TRIM.SP Code for Low-Energy Muon and Ion Implantation

We describe the staged modernization of a local multilayer TRIM.SP implementation, denoted TRIM.SP-NL. The work included behavior-preserving refactoring, correction of localized bookkeeping and initialization defects, support for more than five elements per layer, reproducible build modes, and a runtime-selectable random-number-generator (RNG) interface. The RANLUX RNG remains the default reference backend, while PCG32 and xoshiro256** provide faster alternatives without changing the input-deck format. The code validation combined exact regression testing with statistical comparisons over 13 representative target multilayer configurations, comprising 143 energy-configuration points per backend and 429 simulations in total. Across the test suite, PCG32 and xoshiro256** reduced runtime by factors of 1.98 and 2.01, respectively. The implanted, backscattered, and transmitted fractions, as well as the mean implantation depths, range straggling, and representative depth profiles showed no systematic RNG dependence. Together with a Node.js-based graphical interface for setup, scans, and result collection, the Fortran engine forms the open-source TRIM.SP-NL Workbench.
Read the full abstract

Machine Learning

Local Reference Geometry Residual Augmentation for Imbalanced Time Series Classification

Imbalanced time series classification is often addressed by changing the training distribution, objective, logits, or final threshold. These interventions address important biases, yet leave a representation-level question unmeasured: after minority support is reduced, does a learned feature space remain locally reliable around minority regions? We identify a training-local geometry failure: under imbalance, minority cases can lie in sparse, rest-dominated, or mixed feature-space neighborhoods, even when the representation retains useful global class structure. To diagnose and repair this failure, we propose Local Reference Geometry (LRG), a lightweight post-hoc feature augmentation module applied between a fixed feature extractor and the classifier head. Using training features only, LRG measures local exposure and class-mixture risk, then augments each fixed feature with a standardized signed displacement from nearby training geometry and an LDA-projected residual summary. On controlled UCR/Bake Off Redux imbalance benchmarks, paired raw-versus-LRG comparisons show gains for learned, pretrained, and fixed representations, including when LRG is combined with training-level interventions and post-encoder classifier corrections. Ablations show that the gain comes from the signed local residual appended to the original feature, rather than from generic prototype distances, affinity features, scalar statistics, or VLAD-style codes. Further analyses support the proposed local-geometry failure hypothesis: minority neighborhoods become increasingly rest-exposed under imbalance, training-local risk identifies error-prone regions, and LRG gains concentrate in those high-risk regions.
Read the full abstract

All authors Chuanhang Qiu, Yanran Xu, Yue Wang, Anthony Bagnall

Machine Learning

Safin-1: Safety from Within through Memory-Native State Evolution

Long-horizon complex tasks require foundation models to accumulate information, maintain internal states, and adapt over extended interactions. Safety should be an intrinsic property of the model itself, rather than a behavioral constraint relying solely on external safeguards or post-hoc alignment such as supervised fine-tuning. This motivates Safety from Within, where safety-relevant capabilities are represented and invoked through the model's native computation. We present Safin-1, a family of foundation models realizing this principle through memory routing and state evolution. Safin-1 is built on Memory-Anchor Routing across Context History (MARCH), a network architecture that maintains structured memory states and selectively retrieves relevant historical information through content-conditioned routing. It supports test-time adaptation of persistent capability states without repeatedly modifying the backbone, enabling controlled specialization over a shared foundation. We investigate this interface on downstream safety tasks through a Safety State, demonstrating effective state-based adaptation with substantial safety improvements. More broadly, the routed-state interface unifies contextual memory and persistent capability adaptation within the model's native computation, reframing memory from a passive record of prior context into an active substrate for maintaining and evolving model behavior. Evaluations across general capabilities, long-context understanding, retrieval, and efficiency further validate Safin-1. These findings provide a path toward safety as a state-native and adaptively maintainable capability. This work is only an initial architectural exploration of Safety from Within, and substantial further work is needed to realize this broader vision.
Read the full abstract

All authors Ming Zhang, Kaisen Yang, Shu Yu, Ermo Hua, Zhekai Chen, Cheng Jin, Jingnan Zheng, Yi Zhang, Zhongtian Ma, Jiawei Zhou, Sirui Chen, Qiaosheng Zhang, Xiang Wang, Ning Ding, Xia Hu, Bowen Zhou, Youbang Sun, Chaochao Lu

Numerical Analysis

A coercive space-time variational approach to fractional diffusion problems

We consider a fractional diffusion problem with temporal nonlocality acting on the diffusive flux. A coercive space–time variational formulation in Bochner-valued fractional Sobolev spaces is derived and the existence, uniqueness, and regularity of solutions are established. We further develop a conforming tensor-product Galerkin discretization and prove quasi-optimal error estimates in the anisotropic energy norm and improved convergence rates in weaker norms using duality arguments. In contrast to some space-time formulations for classical diffusion, the method preserves the causal structure of the evolution problem and leads to a time-stepping procedure with memory terms. On uniform time grids, the discrete history operator has a lower-triangular Toeplitz structure which enables an efficient implementation using fast recursive convolution techniques.
Read the full abstract

Machine Learning

Assessing Alignment and Stability of Feature Importance Explanations via Weight of Evidence

Feature importance Methods (FIMs) are widely used in Explainable AI to interpret model predictions, yet attribution scores alone often provide limited insight into the underlying reasoning process. In this work, we introduce a novel perspective by embedding FIMs within a hypothesis-testing framework based on Weight of Evidence (WoE). We quantify how strongly the observed evidence supports any given hypothesis on feature importance. The reference hypothesis can stem from domain knowledge, ground truth, or be derived from the FIM itself. This formulation enables a principled evaluation of FIMs, capturing both their alignment with prior knowledge and their variability. We further provide theoretical results linking WoE to attribution variance. Empirical results shows the applicability and flexibility of our strategy analyzing LIME and SHAP explanations in settings with different reference hypotheses. Overall, our framework offers a complementary tool for assessing FIMs through a contrastive, evidence-based lens.
Read the full abstract

All authors Eddie Conti, Claudio Daka, Álvaro Parafita, Antonio L. Alfeo, Axel Brando, Mario G.C.A. Cimino

Machine Learning

Foundation models for electricity price forecasting and battery arbitrage: Can they replace market-specific forecasting models?

Foundation models promise accurate forecasts with little or no task-specific training, but whether they can replace models designed specifically for electricity price forecasting remains unclear. We compare nine variants from five foundation model families, evaluated in zero-shot mode, with two state-of-the-art electricity price forecasting benchmarks in Germany, Poland, and Spain over 2021-2025. Their performance is assessed in terms of point and probabilistic forecasting accuracy, as well as economic value in battery energy storage arbitrage. Only the TabPFN models consistently and significantly outperform the benchmarks across all three markets and all statistical measures. However, this statistical dominance does not translate directly into economic dominance: TabPFN performs best under unlimited bids and riskier quantile-based strategies, whereas the Distributional Deep Neural Network benchmark is more profitable when risk tolerance is lower. Thus, foundation models cannot universally replace market-specific models, and their value depends on both model architecture and the decision problem.
Read the full abstract

Software Engineering

Commit-first LLM judging inherits the judge's own errors

LLM judges, models that score another system's output, can be gamed by the systems they score. Recent work identifies one defence that works: the judge solves the task itself first and commits to that answer, then accepts a candidate only if the two match. We call this commit-first judging, and ask whether shipped software implements it, and what it costs. We audit the default judge configurations of eight widely used evaluation frameworks. Of the 24 configurations in scope, none implement it. Nine implement a variant the literature measures as ineffective, and share one ancestor prompt, traceable through a copied typographical error. In a controlled experiment, an ordinary best-of-N search with no access to correct answers optimises code against one of these configurations, used exactly as documented. On an interval merging task the judge accepted 90 of 96 candidates in one seed and 93 of 96 in the other; every accepted candidate passed every test the search could see and failed a held-out suite it could not. The judge identified the defective line and cited it as grounds for a perfect score. Commit-first judging removed the effect: 0 of 96 in both seeds. On a second task it made matters worse in both seeds: the judge's committed answer was wrong, and in one seed the population converged on it. This is our main finding. Commit-first judging does not remove the anchor that gets gamed, it moves it from the candidate to the judge's own answer, so evaluation is only as good as the judge is at the task. That precondition is cheap to measure in advance, and is task local rather than scale dependent: a smaller judge solved a task the frontier judge failed and resisted gaming where it did not. We also validate our own instruments: five of fifteen claims in our criteria were wrong against verbatim sources, and two held-out checks were unjustified by their specifications.
Read the full abstract

Classical Analysis and ODEs

Translational tiles without spectra in finite abelian p-groups

We construct explicit translational tiles without spectra in three finite abelian $p$-groups. The first is a $64$-point subset of $\Z_4^4\times\Z_2^2$. The other two are a $512$-point subset of $\F_2^{13}$ and a $2187$-point subset of $\F_3^9$. Consequently, the tile-to-spectral implication fails for finite abelian $p$-groups, and it already fails within the class of elementary abelian groups for both $p=2$ and $p=3$. Two elementary mechanisms organize the examples. A two-layer obstruction turns a spectral non-tile with two suitable tiling complements into a tile without a spectrum. A fiber–clique obstruction converts a family of tiling complements with controlled common Fourier zeros into an elementary abelian counterexample. All coordinate data are included. The finite claims are certified by three short, self-contained programs using exact integer arithmetic and exhaustive searches; the accompanying source files recompute every assertion used in the proofs.
Read the full abstract

Computation and Language

Retrieval, Scoring, and Decoding Shape Performance and Stability in LLM-based Conversational Recommendation

Large language models (LLMs) are increasingly used as rerankers in conversational recommender systems, yet measured gains depend strongly on the retrieval and inference protocol. On the ReDial conversational movie recommendation benchmark, we compare proprietary, open-weight, and fine-tuned LLM rerankers with collaborative-filtering and sequential baselines in a shared retrieve-then-rerank pipeline. We vary candidate-pool size, first-stage retriever, and decoding temperature. With a shared semantic top-250 candidate pool and strict candidate-aware scoring, the best proprietary reranker reaches NDCG@10 of 0.1497, compared with 0.0939 for the strongest non-LLM baseline. The same reranker reaches 0.2925 in zero-shot generation, showing that unconstrained scoring can yield a much larger apparent advantage than matched-pool evaluation. No evaluated open-weight LLM outperforms the tuned shallow autoencoder baseline under this protocol. For the strongest proprietary and open-weight rerankers, switching from semantic to collaborative-filtering candidates raises NDCG@10 by more than 50%, showing that measured reranker performance is highly sensitive to candidate generation. For the best proprietary reranker, raising temperature from 0 to 1.0 increases top-10 Jaccard distance from 0.0900 to 0.1240 while mean NDCG@10 changes negligibly, whereas weaker LLMs show larger degradation. These ReDial results support treating candidate generation, candidate-pool size, scoring policy, and decoding configuration as required reporting fields rather than implementation details.
Read the full abstract

All authors Ante Kapetanovic, Tomislav Duricic, Andro Mercep, Emanuel Lacic

General Physics

Dark Matter Imprints on Black Hole Shadows and Chaotic Dynamics: A Comparative Study of Halo Profiles around Sgr A and M87

The gravitational influence of dark matter (DM) halos on the strong-field regime around supermassive black holes (SMBHs) remains largely unexplored. In the present work, we try to make a framework of the halo-modified space times for four well-known DM profiles including the Navarro Frenk-White (NFW), Moore, Burkert, and Dehnen-(1,4,3/4) profiles, respectively. We derive the corresponding photon spheres, shadow radii, and innermost stable circular orbits (ISCO) for the Sgr A* and M87*. We then constrain the DM halo parameters by comparing the theoretical values with the observed Event Horizon Telescope (EHT) data, and the robustness was checked by comparing the obtained virial mass of the Milky Way and M87 from the parameters with the values already present in the literature. In addition to this we extend the geodesic analysis by formulating the perturbed orbital dynamics using a Painlevé-Gullstrand Hamiltonian framework, and as a result we compute orbital trajectories, Poincaré sections, and finite-time Lyapunov ex ponents. These chaotic dynamics leave distinct imprints on gravitational waveforms, including phase irregularities and amplitude modulations.
Read the full abstract

All authors Ali Naqi, Mubasher Jamil, Noraiz Tahir, Saira Yamsin

Machine Learning

Stochastic complexity of vectors containing cluster structure

This paper studies the problem of computing the stochastic probability (shortest code length) of the encoded vectors containing cluster structure using Normalized Maximum Likelihood (NML) model. This is of great theoretical and practical importance in data clustering based on Minimum Description Length (MDL) principle, such as for estimating the best number of clusters and best cluster structure for the data. Straightforward computation of the shortest code length of the vector containing cluster structure based on the NML model requires polynomial time with respect to the size of the vector and number of clusters. We show that this is a tractable problem by introducing a recursion formula for the efficient computation of normalizing constant from the NML model. The time complexity of the new formula is linear opposed to previous polynomial time with respect to the size of the vector and number of clusters.
Read the full abstract

Materials Science

Fermi-Level Metal-d Character of Group-6 Bis-Hexahapto Bilayer Graphene

Superconductivity in graphite intercalation compounds requires occupancy of a carbon-derived interlayer band filled by charge from an electropositive intercalant. Bis-hexahapto (eta6-eta6) coordination binds transition metals covalently to two graphene sheets with little charge transfer: the mechanism defining this chemistry removes the donation that superconductivity rests on. We ask, for group-6 Cr, Mo and W in bilayer graphene, whether Fermi-level states retain the interstitial, carbon-p character of ionic references, with bulk CaC6 and same-cell C12Ca/C12Li controls. Electron-phonon coupling is not computed. At one metal per gallery (C12M), the ionic controls retain 68 and 45 per cent of gallery spectral weight in atom-masked interstitial regions at EF; the group-6 systems retain 15, 22 and 23 per cent. The contrast is compositional: group-6 galleries carry more absolute interstitial weight than C12Li, but projected density of states within +/-0.15 eV of EF is carbon dominated in the ionic references and metal dominated (~73 per cent, of which 89-98 per cent is d) in all three group-6 bilayers. Every system studied is metallic; they differ in what carries the Fermi surface. Zone-centre phonons are clean for all three ordered bilayers; the pristine AA bilayer is not, its unstable mode being the AA-to-AB shear. Geometrically, eta6-eta6 coordination requires AA stacking and reverses the intrinsic Bernal preference by 400-600 meV wherever metal is present. Ordered phases are metastable against bulk metal (+2.9-5.0 eV per atom) but bound against isolated atoms (-1.6-4.1 eV); instability ordering (Cr least, W most) matches Cr > Mo > W reactivity. C12Cr is metallic; tungsten alone carries 0.65 muB. Bis-hexahapto intercalation delivers strong interlayer bonding and a definite AA registry, but not the interlayer-band character of superconducting graphite intercalation compounds.
Read the full abstract

Computation and Language

KItCAT: Knowledge Injection via Input Corruption for Auto-regressive Training

LLMs acquire vast amounts of knowledge during pre-training, but often lack the specialized knowledge needed to answer questions from niche sources such as manuals or technical documents unseen during pre-training. Continued pre-training (CPT) is widely used to inject such knowledge into model parameters. However, niche documents seldom repeat facts, making it difficult for CPT to robustly acquire such knowledge. Recent works address this by generating multiple paraphrases of the new knowledge, but paraphrasing is computationally expensive and typically requires powerful LLMs. In this work, we introduce KItCAT: Knowledge Injection via Corrupted Auto-regressive Training, a lightweight training strategy that reduces the need for paraphrasing in decoder-only LLMs. KItCAT augments standard next-token prediction by stochastically corrupting the input sequence. During training, a random subset of input tokens is replaced with other vocabulary tokens while the original next-token labels are kept unchanged. This simple intervention generates diverse training inputs from each sample, enabling large-scale data augmentation at negligible cost. We show that KItCAT consistently improves over CPT across multiple datasets and model families. Code is available at https://github.com/meghanadhpulivarthi/KItCAT.
Read the full abstract

All authors Meghanadh Pulivarthi, Kushagra Bhushan, Vineet Kumar, Gaurav Pandey, Jaydeep Sen, Dinesh Raghu, Sachindra Joshi, Yatin Nandwani

Software Engineering

Framework and Benchmark for Code-Driven Agentic Testing in Web Development

End-to-end GUI testing is essential for verifying web applications, yet existing evaluations rely on predefined checklists and are confined to the data and frameworks of web generation benchmarks, leaving the bug-discovery ability of vision-language models (VLMs) systematically untested. We introduce Code-driven Agentic Testing (CAT), a paradigm in which the agent writes Playwright code to drive the browser, gathers feedback, and autonomously explores web applications to uncover bugs. We instantiate CAT with CATJudge, an agentic framework that unifies Browser-Use and Computer-Use tools within a single environment and CATTest, a benchmark of 102 AI-generated web applications with carefully annotated bugs, built through close human-AI collaboration to feature complex interactions and subtle defects. Experiments with mainstream VLMs show that all evaluated models perform poorly, revealing a clear gap between current VLM capabilities and the demands of real-world testing in AI web development. We release our code and data at https://github.com/SleepyWithoutCoffee/CATJudge.
Read the full abstract

All authors Bin Hong, Zhenchao Zhang, Jiyuan He, Kai Zhang, Zhenya Huang

Group Theory

Generalized Graph Compositions with Applications to Difference Graphs of Finite Groups

The difference graph $D(G)$ of a finite group $G$ is obtained from the edge difference between its intersection power graph and power graph, after deleting isolated vertices. This graph has already been studied, with sufficient conditions for connectedness and a diameter bound $6$ for finite groups satisfying those conditions. We use generalized graph composition to reduce $D(G)$ to a graph $B(G)$ on the cyclic subgroups of $G$, so that connectedness and diameter are determined by the subgroup structure of $G$. We obtain a general criterion for the non-emptiness of $B(G)$ in terms of branching subgroups and b-normality, and characterize its connectedness for finite $p$-groups, non-cyclic finite abelian groups, and non-abelian groups with both trivial and non-trivial center. Combined with the previously established cyclic-group case, this gives a complete characterization of non-emptiness and connectedness of difference graphs for all finite groups. The successive structural cases lead naturally to the sharp diameter bounds $2,3,4,$ and $5$. For centerless non-abelian groups, connectedness is governed either by a unique branching subgroup or by an auxiliary graph $\mathcal A(G)$; in the latter case \[ \operatorname{diam}\mathcal A(G)-1 \leq \operatorname{diam}B(G) \leq \max\{4,\operatorname{diam}\mathcal A(G)+1\}, \] and both bounds are sharp.
Read the full abstract