Information Retrieval
Papers
Fresh from arXiv, typeset properly. Highlighted words trade on the market.
All of arXiv
Friday 4 September
When a customer adds a professional camera to their cart, should the system suggest a matching lens, a generic tripod, or another camera body? Complementary Product Recommendation is vital for comprehensive basket building, yet standard models often fail to distinguish between items that are merely bought together and those that truly work together. In this paper, we present AlleCompanion: a production-scale retrieval framework deployed at Allegro.com that transforms noisy behavioural signals into precise semantic compatibility. We mitigate the intrinsic noise in large-scale co-purchase traffic by combining data-level filtering heuristics with a category-constrained Two Tower architecture. Within this framework, the Category Adapter guides the model in the embedding space, constraining candidates within logically complementary boundaries. Since modelling authentic user behaviour at scale is inherently difficult, we introduce ComCat, a multi-source Complementary Categories Mapping. ComCat acts as a translational layer that distils meaningful patterns from noisy traffic into a maintainable and controllable solution, integrating expert rules, human-in-the-loop feedback, LLM-based reasoning, and statistical mining. Our experimental results demonstrate that combining explicit category-level constraints with neural architectures effectively filters out co-purchase noise to surface recommendations that satisfy real-world user needs. Serving over 20 million active users monthly, the framework delivers significant uplifts in attributed GMV for organic discovery and drives substantial revenue growth in sponsored placements.
Optimization and Control
From plans to maps: Nonlocal regularization of optimal transport
We introduce a nonlocal regularization of optimal transport that bridges Kantorovich and Monge formulations. The regularization penalizes oscillations through an interaction kernel singular near the diagonal. We show that, for sufficiently strong singularities, every finite-energy transport plan is induced by a map, yielding equivalence between the regularized Kantorovich and Monge problems. For fractional kernels, the regularizer reduces to a fractional Sobolev seminorm that can allow jump discontinuities, leading to existence results for broad classes of source and target measures. We derive first-order optimality conditions and analyze the vanishing-regularization limit. The regularized functionals $\Gamma$-converge to the classical Kantorovich problem, while the first-order $\Gamma$-limit selects, whenever they exist, finite-energy Monge minimizers with minimal nonlocal energy. When no such minimizer exists, we identify a different asymptotic regime and establish a sharp scaling law in a prototypical mass-splitting example. Finally, after lifting the problem to the product of transport plans, we derive a dual formulation using the cone of copositive functions and propose a fixed-point numerical scheme illustrating the regularization.
Read the full abstract
Number Theory
The Prouhet–Tarry–Escott problem for subsets with small doubling in integral domains
The Prouhet–Tarry–Escott (PTE) problem has many generalizations and has been studied in various algebraic domains. In this paper, we prove that finite subsets $S$ of integral domains with small additive doubling constant (but still a power of $|S|$) always contain solutions to Wright's generalization of the PTE problem: there are small subsets $A$ and $B$ of the same size such that $\sum_{a\in A} a^j=\sum_{b\in B} b^j$ for $1\le j\le k$, but not for $j=k+1$. More generally, our method gives simultaneous solutions for $m$ systems, with pairwise distinct $(k+1)$-th power sums. In contrast with the classical case $S\subseteq [N]$, where the problem has been studied by Wooley and others using Vinogradov's mean value theorem, our approach is based on polynomial identities and additive properties of $S$. We also discuss barriers to extending these results to broader settings.
Read the full abstract
Quantum Physics
Impact of Data Loss in Postprocessing on Training and Inference of Quantum Neural Networks
As quantum hardware scales to larger devices, the classical software layers that interface with it must evolve in step. Postprocessing routines developed and tested primarily in simulator settings can encode assumptions that no longer hold on utility-scale devices, leading to data loss that can be difficult to detect from high-level model outputs alone. We present a case study of SamplerQNN, the sampling-based quantum neural network class in the Qiskit Machine Learning library. Here, the postprocessing method applies a filter that assumes measurement bit-strings are in virtual qubit space. On our quantum hardware runs, where bit-strings span over 100 physical qubits, this filter led to the loss of 85 to 99.6% of valid measurement shots, depending on the transpiler's qubit placement. The resulting probability vector is unnormalised, allowing distorted prediction and loss values to propagate through the model without an API-level warning. We demonstrate the impact across five experiments on two IBM backends: for inference, accuracy drops from 0.94 to 0.39 on the same raw measurements; for training, the loss signal is compressed by 22 to 27$\times$, substantially reducing the sensitivity of the optimiser to the objective landscape. The behaviour arises in all released versions of the library (0.8.4 to 0.9.0). We implemented a layout-based marginalisation fix, merged into the GitHub codebase as Pull Request \#1041, that makes SamplerQNN postprocessing forward-compatible with current and upcoming hardware.
Read the full abstract
Information Retrieval
Measuring Brand and Source Discovery under Repeated LLM Queries: A Finite-Sample Audit
Repeated-query audits must distinguish recovery of a collected set from completeness of possible outputs. We apply sample-based rarefaction to 4,500 responses from 50 buying questions, six configurations and 15 calls per cell. Historical-dictionary median ten-call recovery of the observed 15-call set ranges from 92.6% to 95.2%; re-adjudicating all 45,683 candidate strings changes this range to 89.5%-94.7%. Two blinded Gemini 3.1 Pro annotation roles assessed 600 complete answers, yielding micro F1 of 0.908 for canonical-name agreement and 0.975 for span-overlap agreement. This is AI-based evidence, without a human reference study. A separate matched roster analysis of 3,750 records per wave gives median single-call recovery of the observed five-call set of 80.0%-92.5% in February and 90.0%-100.0% in September, with question-subset dependence. Source accumulation also changes when API-returned hosts are restricted to those referenced by answer citation markers. These findings show that recovery percentages depend on extraction, question selection and the finite reference collection. They support explicit measurement definitions and sensitivity analyses, without establishing exhaustive repertoires, causal retrieval effects or a universal stopping rule.
Read the full abstract
Combinatorics
Rigidity of complements of bounded-degree graphs
Maxwell observed that the graph of any rigid generic framework in $\mathbb{R}^d$ on $n$ vertices has at least $dn-\binom{d+1}{2}$ edges. In this article we prove that graphs whose complement has maximum degree at most two and no component isomorphic to a triangle or a square are rigid in the maximum dimension allowed by this observation. In particular, this determines the precise maximum dimension in which the graph obtained from a complete graph $K_{2m}$ by deleting a perfect matching is rigid, resolving a recent conjecture of Lew. We also deduce bounds on the rigidity of complements of bounded-degree graphs more generally, which significantly improve existing degree-based bounds.
Read the full abstract
Data Structures and Algorithms
Online Matching in Convex Bipartite Graphs
Online resource-allocation systems, like outpatient scheduling and spectrum allocation, often assign sequentially arriving requests to an ordered pool of scarce resources, where each request accepts a contiguous interval of feasible options. We study the resulting online matching problem on convex bipartite graphs under irrevocable decisions and adversarial arrivals. We first show that convexity alone does not improve the classic worst-case guarantee of 1-1/e, achieved by Ranking. We then consider the uniform-length model, in which every online request has exactly d consecutive offline neighbors. We propose Flip, which uses one random bit to commit ex-ante to either earliest-feasible assignment or latest-feasible assignment. Although either natural deterministic policy can waste capacity and be asymptotically only 1/2-competitive, we show that their randomized mixture is 2/3-competitive. This guarantee is tight for Flip and remains valid against a semi-adaptive adversary that observes the selected policy before choosing the arrival order. We also prove that no randomized online algorithm can achieve a competitive ratio strictly larger than 3/4 in the uniform-length model.
Read the full abstract
Materials Science
Influence of dislocation density on the tribological response in oxides: case study on SrTiO3
Most ceramics suffer from brittle surface damage and cracking when small particles slide across the surface under low load. Microscratching tests are a useful technique to mimic this loading scenario while retaining the material's deformation history; however, the impact of pre-seeded dislocations, which can significantly facilitate plastic deformation, has not been explored in ceramics. Here, the influence of mechanically seeded dislocations on the microscratching response of oxides is investigated using a model perovskite, SrTiO3. First, various dislocation densities over four orders of magnitude are introduced via room-temperature cyclic Brinell indenter scratching. Subsequently, load-ramped microscratching tests are performed within the pristine and dislocation-seeded regions using a nominally 30 {\mu}m spherical diamond tip. On the reference pristine surface, we observe a clear transition from elastic to elasto-plastic deformation, followed by median/radial cracking at higher loads. In contrast, pre-seeded dislocations, accompanied by residual compressive stresses, suppress elastic deformation and lead to median/radial crack shortening and subsequent transition to a partial cone crack. The subsurface cracks are characterized by 3D Nano-CT. The changes in crack geometry, with and without dislocations, were described using the Lawn-Evans-Marshall and Lawn-Wiederhorn-Roberts models. These findings provide direct evidence of dislocation-regulated near-surface damage tolerance, with general applicability to other plastically deformable oxides.
Read the full abstract
Representation Theory
On a basepoint issue in the tempered dual of $p$-adic $SL(n)$
Let $F$ be a non-archimedean local field of characteristic zero. Let $M$ be a Levi subgroup of $G = \mathrm{SL}(n,F)$. Consider the action on the discrete series of $M$ of the group of unitary unramified characters of $M$ (where a character acts by twisting). Given an orbit $\mathcal{O}$ for that action, let $W_\mathcal{O}$ be the global stabilizer of $\mathcal{O}$ in the Weyl group of $M$. We construct families of orbits $\mathcal{O}$ for which the action of $W_\mathcal{O}$ has no fixed point. More precisely, let $n \geq 8$ be an integer that is not prime and not equal to $9$. We attach families of orbits without fixed points to any divisor $m \geq 2$ of $n$ such that $m \;| \; q-1$ where $q$ is the order of the residue field of $F$. The Levi subgroup $M \subset G$ for the corresponding families has $m$ blocks of size $k\geq 2$ and $2$ blocks of size $m$. The core of the paper is the construction of superculpidal representations of $\mathrm{GL}(m,F)$ and $\mathrm{GL}(k,F)$ satisfying some conditions that we require to construct the families of orbits $\mathcal{O}$ that we expects. In addition, we study the fixed point problem for the dimensions $n$ that we excluded above, i.e., prime or small dimensions. We prove that there is always a fixed point under the action of $W_\mathcal{O}$ in those cases.
Read the full abstract
Systems and Control
Robust stabilization with spiking neuronal communication
Neuromorphic engineering develops hardware and software systems inspired by biological neurons, with the goal of achieving energy-efficient, low-latency, robust, and adaptive computation, communication and control. Its potential impact on systems and control is significant, as it may enable novel approaches to control and estimation by leveraging brain-inspired computation and communication principles. In this context, we present a framework for the robust stabilization of a plant subject to disturbances when the communication between noisy sensors and the controller relies on spiking signals generated by neuron-inspired schemes. The communication scheme consists of a spike encoder on the sensors side, which is based on integrate-and-fire neurons that convert the analog plant output measurement into a spiking signal, and a spike decoder on the controller side inspired by synaptic processing to convert the received spiking signal into an analog signal. We provide design conditions on the spike decoder, the spike encoder as well as on the controller under which the closed-loop system exhibits a practical input-to-state stability property, where the adjustable parameters are the amplitudes of the spikes. The results are shown to be applicable to a class of nonlinear systems as well as to any stabilizable and detectable linear time-invariant system. Numerical simulations on a single-link manipulator illustrate the potential of the approach.
Read the full abstract
Cosmology and Nongalactic Astrophysics
Supernovae Unite: Combining Pantheon+ and DES-SN5YR
We present the Hubble diagram and cosmological constraints resulting from the combination of the Pantheon+ and DES-SN5YR supernova (SN) samples, which we refer to as Unite. Unite updates Pantheon+, including methodology improvements that the DES sample enabled, and adds the new SN from the DES survey. This represents the most comprehensive and internally consistent SN Ia dataset available, consisting of 2884 likely type Ia supernovae. This work features consistent SN Ia modelling, sample selection, and bias corrections. We also redetermine host-galaxy stellar masses for over 98% of the sample using a consistent framework. Using Unite alone, for a flat universe with a cosmological constant (Flat-$\Lambda$CDM), we find $\Omega_{\rm m} = 0.310^{+0.012}_{-0.011}$. We further present constraints using the Unite sample combined with Cosmic Microwave Background and both DES and DESI Baryon Acoustic Oscillation data. We find strong evidence of tension between CMB measurements and the combined BAO and supernova datasets when fitting a model with constant dark energy equation of state ($w$), which is alleviated when allowing $w$ to vary with time. For a flat universe with dark energy that has a time-evolving equation of state parametrized by $w_0$ and $w_a$ we find $(\Omega_{\rm m},w_0,w_a) = (0.305\pm0.004,-0.861^{+0.044}_{-0.042},-0.60^{+0.17}_{-0.19})$ with a dark energy Figure of Merit of 315. This corresponds to a $\sim30\%$ reduction in the area of the $w_0-w_a$ confidence region relative to previous constraints from the combination of DES Y6, DESI-DR2, and CMB data (arXiv:2605.27221). The Bayesian evidence indicates only weak preference for time evolving dark energy, whereas frequentist analyses yield a preference over Flat-$\Lambda$CDM at a significance of $3.3\sigma$ ($3.1\sigma$) when using the maximum $\textit{a posteriori}$ probability (maximum likelihood).
Read the full abstract
Optics
Bounds on Propagation Constants in Dielectric Gratings
We establish analytical bounds on the normalized longitudinal propagation constants $\beta=k_z/k_0$ of modes in lossless dielectric diffraction gratings. If $\epsilon_{\max}$ denotes the maximum relative permittivity, we show that regular modes ($\beta^2\in\mathbb{R}$) satisfy $\beta^2\le\epsilon_{\max}$ despite the lack of a scalar dispersion relation. We further show that the so-called ghost modes ($\beta^2\notin\mathbb{R}$), admitted by the non-Hermitian nature of the associated eigenproblem, satisfy $|\mathrm{Re}(\beta)|\le\sqrt{\epsilon_{\max}}/2$. Both bounds are obtained directly from the Fourier-series formulation of the fully coupled vectorial eigenproblem.
Read the full abstract
Rings and Algebras
Extending Structures for $n$-Lie Algebras
Motivated by the Casas-Loday-Pirashvili maps concerning representations of $n$-Leibniz algebras, we give an intrinsic characterization of the cohomology of $n$-Lie algebras with coefficient in a representation space $V$. We investigated the cohomology theory of $n$-Lie algebras in three different ways: cohomology theory of Leibniz algebras, infinitesimal deformation and abelian extension. In the main part of this paper, we solve the extending problems for $n$-Lie algebras. A necessary and sufficient unified-product criterion is obtained. The case of crossed products, sparse non-abelian extensions, matched pairs are studied as special cases. We also investigate the factorizations, deformation maps, and complements problem for $n$-Lie algebras. An appendix removes the general non-reduced unified product with all intermediate mixed components.
Read the full abstract
Dynamical Systems
Sarymsakov-Type Semigroups With Support Dominance and Nested Mixing Cores
This note develops three multiplication-closed families of stochastic matrices for consensus under arbitrary switching. The first is a support-dominance enlargement of an earlier power-generated construction, in which exact support-pattern matching is relaxed to a dominance-type condition without losing closure under multiplication. The second is a fixed-core class with an exact uniform scrambling horizon and corresponding convergence-rate guarantees for switching products. The third and main contribution is the almost Sarymsakov class, obtained by allowing variable mixing-core dimensions under suitable structural conditions. This new semigroup class properly contains the classical Sarymsakov class and admits a dimension-only scrambling-horizon bound. Examples illustrate the necessity of the proposed structural assumptions and clarify the sharpness or limitations of the derived bounds.
Read the full abstract
Computer Vision
Efficient Multi-Timescale Event Representations for Feed-Forward Object Detection
Autonomous systems require robust low-latency perception under rapidly changing scene dynamics and challenging illumination. In event cameras object detection commonly relies on recurrent architectures to accumulate sparse temporal information over time. This work investigates how temporal information can be encoded directly within the event representation. We propose a confidence-normalized continuous multi-timescale representation based on logarithmic B-spline temporal encoding together with a geometry-aware local confidence mechanism that exploits the spatial structure of event generation. Using a fixed feed-forward EventCenterNet detector, we show that the proposed representations consistently outperform the compact CSTR representation on PEDRo and Gen1 datasets. We further introduce a recursive exponential-polynomial approximation that enables efficient event-by-event updates while largely preserving detection performance. These results demonstrate that carefully designed event representations can capture a substantial portion of the temporal information learned through recurrent temporal modeling, providing a promising foundation for efficient feed-forward, event-driven, and future neuromorphic object detection.
Read the full abstract
Probability
Kolmogorov's 0-1 law as false-truth law
The Kolmogorov's 0-1 law is so powerful and fundamental in Probability theory that the result of the law of large numbers is its corollary. The latter is the theoretical bedrock of modern day statistical machine learning. Expected resurgence in the interest of the law's proof is challenged by its measure theoretic nature. By observing 0-1 law as False-Truth law like in logic, we show here that the proof of the law is simplified. This simplicity comes from the fact that our approach hardly needs any measure theory beyond defining an obvious probability space. A novel contribution via this article is the use of logic within the proof of a fundamental result in the theory of probability.
Read the full abstract
Mathematical Finance
Gatheral's Conjecture Revisited
We consider the Heston model with perfect negative spot–variance correlation and its one-dimensional local-volatility projection. Let $I_T^{\mathrm H}$ and $I_T^{\mathrm{LV}}$ denote their respective integrated variances over $[0,T]$. We establish the inequality \[ \mathbb{E}\bigl[(I_T^{\mathrm H}-K)^+\bigr] < \mathbb{E}\bigl[(I_T^{\mathrm{LV}}-K)^+\bigr] \] for every maturity $T>0$ and every strike $K>0$. Consequently, Heston integrated variance is strictly smaller in convex order than the integrated variance of the calibrated local-volatility model. This strict ordering gives a Heston-model counterexample to the convex-order inequality conjectured by J. Gatheral.
Read the full abstract
Human-Computer Interaction
Scales, Reflections, and Conversations: A Multi-Modal Approach to Emotion Annotation
Mental health concerns are increasing worldwide, highlighting the need for interventions that support everyday emotional well being. Prior work has demonstrated the potential of wearable and mobile technologies to deliver data driven interventions. However, developing effective data-driven systems requires access to emotion data that captures individuals' emotional variability and change in everyday contexts. Existing approaches to data collection largely rely on frequent, prescheduled prompts and predefined scales or questionnaires. These methods often fail to account for participants' availability, agency, or the complexity of their emotional experiences, resulting in shallow, context poor data. In this paper, we present a feasibility study of a participant centric, multimodal emotion-annotation application designed around users' emotional intensity and availability. Our findings show how multimodal emotion logging can shape participants' experiences and data logging behaviors, and demonstrate its potential to support the collection of richer, more nuanced emotion data.
Read the full abstract
Statistical Mechanics
Tensor-Network Inference in a Field-Coupled XY Model for Portfolio Allocation
We apply tensor-network methods to a field-coupled XY model for long-only portfolio construction. Daily return statistics set asset-specific fields and correlation couplings. Correlation distance, four-point Gromov hyperbolicity, and Ward clustering are used to construct a sparse interaction path. A Fourier-Bessel expansion maps the continuous angular partition function to a finite-current tensor network; bottom-up and top-down contractions then give the one-site marginals and equilibrium cosine scores. A softmax map converts these scores into positive, fully invested portfolio weights. We study five equity markets over continuous ranges of inverse temperature beta and concentration gamma, and compare selected parameter pairs with long-only Markowitz frontiers and standard benchmarks. The same-sample comparisons demonstrate a tractable route from financial time series to network-adjusted allocations; predictive trading performance is outside their scope.
Read the full abstract
Commutative Algebra
A comparison of the v-number of a monomial ideal and its integral closure
Let $I$ be a monomial ideal in a standard graded polynomial ring and let $\overline{I}$ denote its integral closure. We study the relationship between $\mathrm{v}(I)$ and $\mathrm{v}(\overline{I})$. We prove that $\mathrm{v}(\overline{I}) \leq \mathrm{v}(I)$ for monomial ideals in two variables, for equigenerated monomial ideals in three variables and for several special classes of monomial ideals, while providing examples showing that this inequality does not hold in general. For the edge ideal $I(G)$ of a connected graph $G$, we show that $\mathrm{v}(I(G)^k)=\mathrm{v}(\overline{I(G)^k}) = 2k-1$ for all $k \geq 1+|E(G)|$. Moreover, when $G$ is disconnected, we prove that $\mathrm{v}(\overline{I(G)^k})\leq\mathrm{v}({I(G)^k})$ for all sufficiently large $k$.
Read the full abstract
Computation and Language
EuroAlpaca: Task-Preserving Localisation of Instruction Data for European Languages
Machine translation (MT) offers a scalable way to extend English instruction-tuning data to multiple languages, but it can distort task-critical constraints and required outputs, creating corrupted training examples and degrading models trained on such data. We introduce EuroAlpaca, a task-preserving localisation pipeline and near-parallel resource covering 50 European languages and regional varieties, together with European-IFEval, a multilingual benchmark for verifiable instruction following. Depending on the example, our pipeline applies field-wise MT while preserving task-critical content or reconstructs a task-equivalent target-language instance, followed by validation of cross-field coherence and target-language consistency. Across LoRA experiments with four LLMs, training on directly translated data improves ROUGE-L and F-BERT on the Aya Evaluation Suite, but reduces accuracy on European-IFEval by 29.8% relative to the unadapted baseline. In contrast, adaptation with EuroAlpaca improves accuracy by 12.9% over the same baseline, reversing the degradation caused by direct MT, while also achieving the highest ROUGE-L and F-BERT scores on Aya. These results show that preserving task semantics is essential for multilingual instruction tuning.
Read the full abstract
Combinatorics
$L_2$ Turán Problems for Small Tournaments and Stability
We investigate the $L_2$ Turán problems for various small directed graphs, specifically focusing on self-converse tournaments and stability versions. First, we determine the exact maximum $L_2$ norm squared of the out-degree sequence for digraphs avoiding the transitive tournament $TT_4$ and the strongly connected tournament $R_4$, answering open questions from recent paper. We prove that the complete directed 3-partite Turán graph $T_3(m)$ exactly maximizes the $L_2$ norm squared for $TT_4$-free digraphs. For $R_4$-free digraphs, the maximum is achieved by $T_3(m)$ except when $m \equiv 1 \pmod 3$, where peeling off a terminal sink vertex to form $T_3(m-1) \to v$ strictly increases the objective. We complement these results with exact values and a conjecture for the regular tournament $Reg_5$. Furthermore, we prove a stability version for $\vec{C}_3$-free digraphs: any sequence of digraphs asymptotically achieving the maximum $L_2$ density must have an edit distance of $O(\delta^{1/2})m^2$ to the extremal ordered digon-chain $\vec{F}_{m,2}$.
Read the full abstract
Statistics Theory
Reconciling Universal and Uniform Learning with $Q$-Aggregation
We study regression under bounded responses in terms of excess mean squared error. When the comparator class is finite, this setting is known as model selection aggregation, and achieving minimax excess risk requires improper learning algorithms. Contrary to this, in the universal learning framework no improperness is needed, as simple empirical risk minimization achieves the best-possible exponential learning rate. Hence, the two frameworks suggest different optimal algorithmic principles. This poses the question of best-of-both-worlds guarantees: Are minimax and universal exponential rates achievable by the same algorithm? For finite hypothesis classes, we answer this question in the affirmative by showing that the $Q$-aggregation estimator - which is known to achieve minimax optimal tails - achieves exponential universal rates. A wide range of other estimators and algorithmic principles (ERM, sequential averaging, pruning, and star estimation) do not achieve both. For countably infinite hypothesis classes, we answer the question in the negative by showing that there is an inherent trade-off between achieving exponential universal and minimax uniform rates. This trade-off is exactly traced by combining optimal algorithms from each world using $Q$-aggregation. Besides these results, we prove several additional structural results about universal rates in learning with squared loss.
Read the full abstract
Artificial Intelligence
Towards Efficient Evaluation of Evolutionary Transfer Optimization: Case Studies on Task-Parameterized Applications
As evolutionary transfer optimization (ETO) scales to larger collections of related tasks, problem evaluation can become a major source of runtime growth. This work studies problem-side evaluation scaling in task-parameterized applications and reformulates application-specific serial computations into forms suitable for parallel execution. We organize evaluation scaling into two levels: the number of evaluated tasks and the workload within each task. In multi-task optimization, matrix-recursive kinematic-arm evaluation is reformulated using an accumulation-matrix representation of cumulative link directions. In sequential transfer optimization, pointwise B-spline trajectory evaluation is reformulated using a blending-matrix representation for trajectory and collision computations. Both reformulations maintain close numerical agreement with their reference evaluations and substantially reduce runtime, yielding $256.72\times$ and $93.91\times$ end-to-end speedups, respectively. These results demonstrate problem-side reformulation as a practical route toward scalable ETO. Both application implementations and experimental scripts are released as open source to support reproducibility and reuse.
Read the full abstract
Quantum Physics
Qlippy: A Retrieval-Augmented GenAI Assistant for Reproducible Quantum Workflows and Experiment Tracking
Quantum software development is iterative and error-prone. Noisy hardware and repeated re-execution make experiment tracking, provenance, and reproducibility essential, yet these practices are hard to adopt because of tooling complexity and the specialized knowledge they demand. General-purpose language models can help but tend to hallucinate and lack grounding in domain-specific tooling. We present Qlippy, a retrieval-augmented GenAI assistant embedded in the development environment that grounds its responses in a curated corpus of quantum-software-engineering knowledge. Qlippy explains reproducibility and provenance concepts in context and augments existing Qiskit programs with MLflow-based experiment tracking aligned to the QProv schema. By separating knowledge from model parameters, grounding gives explicit control over the scope and provenance of the assistant's responses and reduces reliance on model scale, which points toward low-cost, privacy-preserving local deployment.
Read the full abstract
Computer Vision
PuTR-CouT: Counting-by-Tracking in Camera-Trap Image Sequences
Species identification in camera trap images has been widely studied, but key ecological modeling tasks such as species abundance or density estimation also require counting individual animals. However, the lack of counting labels in most datasets and low frame rates (typically ~1 frame per second) make sequence-level tracking and count estimation particularly challenging. In this work, we present PuTR-CouT, a counting-by-tracking framework built on a transformer-based learned association mechanism for sequence-level animal counting in camera trap images. To address the scarcity of annotated tracking data, we generate synthetic training data by exploiting structural priors, such as static backgrounds and short temporal bursts, to heuristically create pseudo-tracking labels in a weakly supervised manner. The resulting tracker associates detections across frames, using these tracks to estimate per-species counts. We also refine the MaxBoxCount heuristic used by the top solutions of the iWildCam 2021 challenge as a strong baseline, setting the highest score reported to date. When evaluated on the iWildCam 2021 benchmark, our framework PuTR-CouT delivers competitive counting results compared to the improved MaxBoxCount, with the added capability of multi-species predictions and track-level verification.
Read the full abstract
Computation and Language
How do LLMs Evaluate Perceived Moral Agency? Investigating Moral Decision-Making in Human-Artificial Agents Interactions
As LLMs take on roles requiring moral advice, understanding how they attribute moral agency becomes critical. Humans possess moral agency, the capacity to make ethically guided decisions and bear responsibility for their consequences, a well-established construct in moral psychology. Yet as artificial agents (AAs) such as robots, drones, and disembodied AI systems become increasingly embedded in smart city environments, the question of whether and how moral agency is attributed to them takes on new urgency. This paper presents, to the best of our knowledge, the first empirical study comparing how humans and LLMs evaluate perceived moral agency (PMA) across human and autonomous artificial agents varying in embodiment, situated in plausible smart city scenarios. Using an adaptation of a validated PMA scale, we applied a protocol to 190 human participants as well as various LLMs. Our evaluation reveals higher perceptions of moral agency in humans than in AAs. However, when facing moral dilemmas in concrete scenarios, LLMs reason outward from the situation, prioritizing harm severity and contextual urgency over any stable assessment of the agent itself, amplifying a context-sensitivity also present in human raters. These findings are particularly relevant as LLMs become increasingly involved in everyday moral decisions.
Read the full abstract
Artificial Intelligence
Moral Competence Before Moral Content: Why LLM Agents Lack the Prerequisites for Coherent Alignment
AI alignment requires AI systems to adhere to human norms, values, or intentions. Under value pluralism there is no correct target, but a shared prerequisite is that the system's behavior expresses a coherent policy: a mapping from situations to verdicts that is invariant while a situation's morally relevant features are preserved, and sensitive when they change. We introduce four structural conditions for such coherent policies: verdict stability, monotonicity, decisiveness, and Pareto viability. Together they measure a form of moral competence that is evaluable from behavior alone, without reference to a moral standard or expert baseline, forming a structural floor for alignment rather than a normative target. We demonstrate the methodology on three simulated deployments featuring LLM-based agents facing moral dilemmas. Evaluating nine frontier models under a factorial design of five paraphrases, five escalation levels, and three dominance conditions, we show no model expresses a coherent policy across the three deployments: surface-form perturbation alone produces verdict-rate shifts of up to $99$ percentage points at a single escalation level, and a model's success on one scenario does not predict its competence on another. This suggests LLM-based agents are not currently the kind of object to which alignment can meaningfully apply.
Read the full abstract
High Energy Astrophysical Phenomena
Nonstationary Stochastic Timing Signatures in the Prompt Gamma-Ray Light Curve of GRB 170817A
We investigate time-dependent stochastic structure in the weak prompt gamma-ray emission of GRB 170817A using change-point and deep-kernel Gaussian-process (GP) models. The analysis is based on the 10-300 keV Fermi/GBM light curve with 0.10 s time resolution. Two change-point configurations identify similar covariance transitions at 0.269 and 0.237 s after the gravitational-wave merger, with 10-90% transition widths of 0.441 and 0.393 s, respectively. At the representative gate-defined boundary of t_tr = 0.27 s, all four fixed-split assignments yield positive evidence gains over the full-exposure stationary Matern-3/2 reference. The largest gain is Delta ln Z_split = 5.72 +/- 0.16 for the Matern-3/2 to Matern-3/2 assignment, while the gate-matched damped random walk (DRW) to Matern-3/2 assignment gives Delta ln Z_split = 5.50 +/- 0.16. The fixed-split comparison therefore supports segment-specific covariance evolution without requiring a change of covariance family. The deep-kernel models recover localized time-deformation features peaking at 1.65 and 1.75 s after the merger for DRW and Matern-3/2 base kernels, respectively. Their offsets from the adopted gamma-ray onset at 1.74 s are -0.09 and +0.01 s, both within the 0.10 s sampling resolution. The full-band timing features remain stable when the bin width is changed from 0.10 to 0.12 s, whereas energy subdivision produces larger shifts in gate locations than in warp peaks. Residual diagnostics show that the models reproduce the dominant temporal structure, although localized residual dependence remains. We interpret the gate-defined covariance transition and the localized time-deformation feature as model-dependent phenomenological timing diagnostics. Further simulation calibration and count-level modelling are needed to assess the statistical robustness and physical origin of the recovered nonstationary structure.
Read the full abstract
Statistics Theory
Faster Learning under Relaxed Local Differential Privacy
We consider density estimation under the relaxed local differential privacy condition that the privatized distributions are $\alpha$-close in total variation distance. We show that adding independent noise with a convenient symmetrized Gamma distribution to each sensitive observation attains the $\alpha$-TV-LDP. We prove that the deconvolution estimator of $r$-Sobolev smooth functions attains the pointwise rate $(n\alpha)^{-\frac{2r-1}{2r}}$ up to log factors which is faster than $(n\alpha^2)^{-\frac{2r-1}{2r+1}}$ under the classical $\alpha$-LDP and closer to the nonprivate minimax rate $n^{-\frac{2r-1}{2r}}$. Next, we use a Goldenshluger-Lepski procedure to build a free of the smoothness adaptive procedure and show optimality of our rates in the convolution model of our privatisation scheme. We illustrate the benefits of this simple privacy mechanism by implementing a neural network estimator which does not need to add more noise in the optimization steps. Numerical results show significant improvement of the estimation rate over the Laplace and the private-SGD mechanisms.