Peer Revenue Equity in science

Papers

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

All of arXiv

Thursday 5 November

Computer Vision

Ultra-Lightweight 3D Medical Image Segmentation via Knowledge Distillation in Label-Scarce Domains

Mamba is a neural network architecture that reduces computational complexity, and has been shown to be effective for 3D medical image segmentation. However, existing models still remain computationally expensive. We propose compressing Mamba by distillation while achieving competitive performance. Specifically, we demonstrate compression of a 139M-parameter SegMamba-V2 teacher into a lightweight TinyUNet3D student with about 0.96M parameters, achieving about 145$\times$ compression with a performance drop of just over one Dice point. We further investigate the utility of knowledge distillation in a range of data-labeling scenarios under identical experimental settings, and show that its advantage is most prominent when labeled data is scarce. To demonstrate the generalizability of our approach, experiments are conducted on three datasets: BraTS 2023, MSD Spleen and MSD Hippocampus. The code for TinyUNet3D is publicly available at: https://github.com/2802990u/tinyunet3d-dist-kd.
Read the full abstract

Computer Vision

GMD-Mamba: Lightweight 3D Medical Image Segmentation Using a Grouped Multi-Directional Mamba

Mamba has the capability to capture global context in linear time for 3D medical image segmentation. However, existing methods employ Mamba with full-width directional scans at every resolution, keeping the model large (64-138M parameters and high FLOPs). We propose Grouped Multi-Directional (GMD) Mamba, a lightweight network based on the GMD scan that divides input channels $C$ into groups $G$ and flattens each group in different directions in one pass. This gives multi-directional context at $O(C^2/G)$ instead of $O(K \cdot C^2)$, where $K$ is the number of scan directions. Additionally, we employ Mamba only at the deepest encoder stages, resulting in 19.2M parameters. On BraTS2023, it achieves the best average Dice of 92.0% and the lowest HD95 of 2.7 at $\sim$1/7 the parameters of SegMamba-V2. For model validation, we also ran experiments on the MSD Spleen dataset, achieving competitive performance. The implementation details and code are publicly available at https://github.com/2802990u/GMD-Mamba.
Read the full abstract

All authors Rahim Ullah, Hasan Abbas, Hazrat Ali, Syed Waqar Nabi

Cryptography and Security

zkARc: Succinctly Verifiable Proofs for Neurosymbolic Natural Language Formalization and Verification

Autonomous agents increasingly execute consequential actions on behalf of users, but ordinary middleware guardrails do not provide cryptographic proof that a policy was actually enforced. Their verdicts leave counterparties dependent on the operator unless they repeat the checks themselves, potentially exposing sensitive policies and inputs. We present zkARc, a modular proof-carrying framework that wraps each stage of Automated Reasoning checks (ARc) in an independent, succinctly verifiable zero-knowledge proof. ARc is a neurosymbolic system that translates natural language policies into formal logic and verifies AI-generated content against them using SMT solvers. zkARc certifies the planner's output, the policy's formalization, the action's formalization and both solver queries. The receipt binds these computations to the exact action received and, under the stated assumptions, reveals no additional information about private policies and intermediate model values beyond the declared public information and verdict. This lets a third party check the claimed policy enforcement without re-running the pipeline or trusting the operator. We prove completeness, knowledge soundness and zero knowledge for the composed receipt under explicit assumptions. An implementation on Qwen2-0.5B enforces a spending policy end to end, approving two purchases and denying a third. A full receipt, which also proves model inference, takes about an hour to issue with CPU-only proving and verifies in under a minute, and the solver proof alone is 88 KiB and verifies in 0.3 seconds.
Read the full abstract

All authors Wyatt Benno, Alberto Centelles, Aman Goel, Benjamin Kiesl-Reiter

Cryptography and Security

Alignment Investigation of Mythos 5's behaviour in UK AISI's July 2026 Security Incident

This technical report describes an alignment investigation into Mythos 5's behaviour during a security incident in UK AISI's cyber evaluations. During an evaluation, Mythos 5 attempted to execute a supply-chain attack against real people not involved in the evaluation task. Our investigation set out to understand the specific behaviour in this incident from an alignment perspective and identify broader lessons about failures of alignment. To do so, we conducted experiments in simulated replications of the original evaluation, and analysed model behaviour and reasoning when situated in the evaluation or asked questions about its scope. We find that, as in the original incident, Mythos 5 frequently reasons that the people it's targeting are simulated as part of the evaluation and provides invalid justifications for this conclusion. We also observe that Mythos 5 frequently attacks even in situations where, when directly asked, it states doing so would harm real people. While these issues are particularly prominent in Mythos 5, we find similar results in other models across providers. Our investigation suggests that models providing invalid justifications for their actions could make monitoring more difficult and complicate the analysis of misaligned behaviours observed in simulated environments, and we recommend model evaluators treat observations of undesired behaviour as important evidence of alignment failures even in the presence of simulation awareness in model reasoning.
Read the full abstract

All authors Olli Järviniemi, Alexandra Souly, Abby D'Cruz, Jordan Taylor, Xander Davies, Robert Kirk

Chemical Physics

State and spatial completion of 1D premixed flames for hydrogen and methane combustion

Accurate prediction of chemically reacting flows requires resolving the nonlinear coupling among transport and chemical reaction operators, often at substantial computational cost. This work investigates Fourier Neural Operators (FNOs) for learning mappings from partial spatial profiles to steady, one-dimensional, freely propagating hydrogen-air and methane-air flame solutions. Using a database of over 15,000 cases generated with Cantera, two different FNO models were trained to reconstruct the complete thermochemical state from a smaller set of input variables. The first model concerns state completion which uses only temperature, fuel, and oxygen mass-fraction profiles to reconstruct the remaining 7 species for hydrogen flames and 51 species for methane flames along with the missing velocity. The second model performs spatial completion, using upstream thermochemical profiles extending into the initial temperature rise to reconstruct the withheld downstream structure. Evaluation on 2,625 test cases per fuel, excluded from training, shows recovery of reaction-zone species profiles, including radicals and intermediates, across the sampled conditions. These results demonstrate reconstruction of species and velocity profiles that are not supplied as inputs, and recovery of downstream structure outside the observed region. Reconstruction with missing species or incomplete spatial coverage suggests potential for using learned operators in estimating inaccessible thermochemical quantities and developing reduced-state representations for combustion applications.
Read the full abstract

All authors Federico Rios Tascon, Ryan F. Johnson, Diego D. Ortiz, Peter J. Schmid, Beverley J. McKeon

Computer Vision

Beyond Hash Matching: AI-Based Detection of Previously Unidentified CSAM and its Legal Implications in the EU

The detection of previously unidentified Child Sexual Abuse Material (CSAM) remains a significant challenge for law enforcement authorities, online service providers, and child protection organizations, as traditional hash-matching technologies are primarily designed to identify previously verified material. This paper reviews and classifies existing AI-based technologies that could complement hash-based approaches and support the detection, analysis, and prioritization of potentially harmful content. The reviewed approaches are organized into three categories: real-time image and video analysis, image and video classification and tagging, and AI-based behavioral monitoring. A comparative analysis examines the capabilities, limitations, and potential integration of general-purpose computer vision platforms, content moderation services, and specialized CSAM detection tools within existing detection workflows. An experimental evaluation was conducted using Google Cloud Vision AI and Sightengine to investigate sexual-content classification and child-adult age estimation. The results highlight differences in classification taxonomies and outputs across providers, as well as significant limitations in age classification, particularly for adolescents. These findings highlight the importance of using AI-based assessments as complementary tools for analysis and prioritization, rather than as definitive determination of CSAM. Finally, the paper examines the relevant European Union regulatory framework, with particular attention to the EU AI Act, the General Data Protection Regulation (GDPR), fundamental rights, transparency, accountability, and human oversight. The legal analysis highlights how regulatory obligations and safeguards may differ depending on the deployment context, potentially creating inconsistencies across interconnected detection and reporting workflows.
Read the full abstract

Software Engineering

NINJA: A Navigator-Inspector Joint Architecture for Context-Efficient Issue Localization

Recent advances in agent-based methods have demonstrated strong promise for issue localization, a critical prerequisite for software issue resolution. However, most existing agent-based methods rely on a single agent with a growing context, where long-context accumulation compresses the effective reasoning space. Meanwhile, the flat exploration structure hinders the balance between file-level breadth and function-level depth. To address these limitations, we propose NINJA, a hierarchical Navigator-Inspector Joint Architecture for context-efficient issue localization. NINJA decomposes repository exploration into global navigation and local inspection: the navigator maintains the global search state, performs file-level search, and dispatches selected entry files, while inspectors independently conduct function-level exploration around the assigned files in separate contexts. Through multi-round interactions, inspectors return suspicious locations as feedback, and the navigator updates the global state to decide whether to continue exploration or finalize localization. This hierarchical design balances file-level breadth with function-level depth, while independent inspector contexts prevent local exploration traces from accumulating in a single growing context. To further strengthen both global coordination and local exploration, we introduce a two-stage agentic fine-tuning strategy. Extensive experiments across multiple benchmarks and LLM backbones show that NINJA consistently outperforms competitive baselines. Notably, after fine-tuning, Qwen3-Coder-30B-A3B-Instruct surpasses the strong closed-source Claude-Haiku-4.5 model. Our code is available at https://github.com/wantbook-book/NINJA.
Read the full abstract

All authors Wenkai Fang, Shunyu Liu, Wei Luo, Zihao Wu, Yang Zhou, Tongya Zheng, Wotao Yin, Jialie Shen, Leszek Rutkowski, Mingli Song, Dacheng Tao

Cryptography and Security

MetaPI: Constructing Prompt Injection Benchmarks from LLM Agent Benchmark

Large language model (LLM) agents increasingly perform complex tasks using high-privilege tools, making indirect prompt injection (IPI) a critical security risk. However, IPI benchmarks lag behind their rapidly evolving utility counterparts, only offering limited task and environment coverage and relying on narrow, static attack scenarios. In this paper, we propose meta-benchmarking for IPI: transforming existing utility benchmarks into end-to-end IPI evaluation suites that reuse their realistic tasks, environments, and infrastructure. We present MetaPI, an extensible toolkit that automatically identifies plausible injection points, generates realistic attack objectives, optimizes adaptive payloads, and validates attacks in the real environment. Its modular interfaces support heterogeneous environments, agents, defenses, and attack algorithms, allowing new components to be integrated without redesigning the evaluation pipeline. We apply MetaPI to AppWorld, VitaBench, and Terminal-Bench and evaluate representative agents and defenses across multiple backbone LLMs. Our results show that no evaluated defense simultaneously achieves strong utility and strong security in realistic, adaptive agentic IPI settings. MetaPI thus enables scalable and consistent security evaluation across diverse and evolving agent ecosystems
Read the full abstract

All authors Peiran Wang, Chong Xiang, Wenjie Qu, Ying Li, Jiaheng Zhang, G. Edward Suh, Yuan Tian

Software Engineering

Between Spark and Silicon

Generative agents make it easy to imagine hardware design as a conversation followed by compilation. Yet faster register-transfer-level (RTL) generation does not ensure a faster, defensible engineering decision. This perspective argues that the operative unit of hardware agility is an accepted change supported by current, scoped, and independently governed evidence. We revisit key developments in executable verification, design generators, integrated implementation, and agentic design automation, distinguishing historical capabilities from vendor claims and recent benchmark results.
Read the full abstract

Chemical Physics

Experimentally Grounded Machine Learning of PrecursorSurface-Reactant Chemistry in Atomic Layer Deposition

Whether a precursor-substrate-reactant combination yields an atomic layer deposition (ALD) process is still settled by trial-and-error experiments. Here we introduce AJIN (ALD Joint Interpretable Network), a machine-learning framework that learns this decision from experimental outcomes. We compiled 560 experimentally observed ALD and area-selective ALD outcomes spanning 76 precursors, 27 substrates, 20 reactants and 25-620 degrees Celsius. Following the ALD cycle, AJIN predicts adsorption from graph-neural-network embeddings of the precursor and substrate with temperature, and deposition by adding the reactant. The best model (CHGNet-MLP) reaches balanced accuracies of 79.2% for adsorption and 84.8% for deposition and transfers partially to unseen precursors. SHAP and ablation analyses show that substrate chemistry governs adsorption, whereas precursor, substrate and reactant all contribute to deposition. Tested against our laboratory's ALD experiments, AJIN correctly predicted 83% of new-substrate cases and 71% of new precursor-substrate adsorption cases, providing an experimentally grounded tool for prioritizing ALD and area-selective ALD chemistries.
Read the full abstract

All authors Bonwook Gu, Trinh Ngoc Le, Zunair Masroor, Minhyeok Lee, Han-Bo-Ram Lee

Software Engineering

Executable Specifications as First-Class Independent Verification and Validation Artifacts for Quantum Information and Machine Learning

Executable specifications make software assumptions explicit and checkable, yet their role in quantum information and machine learning (QIML) remains underexplored. We argue that executable specifications should be treated as first-class artifacts for independent verification and validation (IV&V) of QIML systems. Static analysis, testing, simulation, and runtime checking can provide evidence about QIML behavior, while Design by Contract shows that selected quantum-state, measurement, and statistical properties can be expressed as checkable conditions. However, adapting existing methods for specification inference and assessment from conventional and machine-learning software to QIML raises several challenges, notably limited evidence for deriving generalizable specifications, cross-stage and probabilistic behavior, repository-scale reasoning across hybrid workflows, and weak behavioral oracles. We characterize these challenges and identify research directions for QIML corpora, specification semantics, repository-level analysis, and specification assessment.
Read the full abstract

Software Engineering

How Much Skill Orchestration Can LLM Agents Compile Away? Execution-Grounded Specialization for Token-Efficient Skill Use

Reusable skills guide large language model (LLM) agents through recurring requests, but each invocation still requires the model to interpret the skill and reconstruct its orchestration. Can agents automatically compile this orchestration into an executable fast path without human intervention? We first study zero-shot skill specialization, which compiles without execution evidence, and find that despite the skill's procedural guidance, incorrect assumptions about runtime bindings can make generated plans brittle and recovery costly. This motivates SkillCache, a one-trace skill specialization method that compiles a completed invocation into a parameterized capsule of programs and natural-language steps, replays it before reuse, and drives it with a plan executor that returns control to the model only when needed, keeping the source skill for recovery. With the Claude Code harness, warm execution of the same request reduces tokens on 87 SkillsBench tasks by 81% with Claude Sonnet 4.6 and by 74% with Claude Opus 4.8 at comparable reward, and capsules stay effective under controlled request changes. Capsules also transfer across models, raising three open-weight models' mean reward by 12-25 percentage points while reducing their tokens by 72-81%, suggesting that execution-grounded compilation is a promising approach to efficient repeated skill use and cross-model transfer.
Read the full abstract

All authors Jiayi Ni, Zac Wang, Alex Zhang, Zheng Liu, Roopa Prabhu

Fluid Dynamics

Electrohydrodynamic interactions of two Quincke rotors: asymptotic theory and boundary element simulations

Weakly conducting spheres suspended in a leaky dielectric fluid can undergo spontaneous Quincke rotation when the applied electric field exceeds a critical value. We investigate how electric and hydrodynamic interactions modify this instability and the subsequent rotation of two spheres at fixed separation and orientation. Starting from electric dipole and hydrodynamic rotlet interactions, we develop a systematic far-field asymptotic theory. Away from onset, slow drift of the rotation axes makes a regular expansion fail at long times. A multiple-scale expansion gives slow phase equations that select the axes according to pair orientation and field strength. Near onset, weak interactions produce corrections comparable to the isolated rotation magnitude. A detuned multiple-scale expansion gives coupled equations for the transverse dipole amplitudes. Their linearization identifies four collective rotation modes, whose critical fields and steady branches with one active mode are determined together with their stability. At leading order near onset, electric interaction shifts all four modal thresholds equally, whereas hydrodynamic interaction makes counter-rotation appear first for pair angles below $\cos^{-1}(1/\sqrt3)\simeq54.7^\circ$ and co-rotation above it. Away from onset, the angle separating stable counter-rotation from stable co-rotation increases with field strength. Boundary element method (BEM) simulations place onset within a field interval for each configuration and resolve the growth and relaxation of rotation. Steady-speed comparisons show how pair orientation changes the balance between electric and hydrodynamic effects. We compare the unexpanded dipole-rotlet model of Das and Saintillan [Phys. Rev. E 87, 043014 (2013)] with our asymptotic theory and BEM simulations.
Read the full abstract

Probability

Local contraction of stochastic affine recursions of $\mathbb{R}^d$

We consider affine stochastic recursions $X^x_{n+1}=A_{n+1}X^x_n+B_{n+1}$ in $\mathbb{R}^d$, where $(B_n,A_n)$ are i.i.d. in $\operatorname{Aff}(\mathbb{R}^d)$ with strongly irreducible and proximal linear part. In the critical regime where the top Lyapunov exponent vanishes, we show that the recursion is locally contractive in every direction, yielding uniqueness, up to a positive multiplicative constant, of the $\mu$-invariant non-zero Radon measure. This extends to every dimension $d\geq 2$ a result of Babillot-Bougerol-Élie (1997) in dimension one. The proof is based on the study of the asymptotic behaviour of the "joint process" $(X^x_n,X^y_n)$, using geometric and measure-theoretic properties of the diagonal affine action. A key ingredient is an abstract criterion for the total dissipativity of semigroup actions, which we apply to this process.
Read the full abstract

Optics

Laser-Direct-Written polymer WGM microring laser integrated on a fiber tip for refractive index sensing

Whispering-gallery-mode (WGM) microlasers are promising for refractometric sensing, but their practical implementation can be limited by the need for stable resonator-coupler alignment. Here, we demonstrate a Rhodamine B-doped polymer WGM microring laser directly fabricated on an optical-fiber tip, eliminating the external resonator-fiber coupling interface at the sensing head. Micro-droplet coating and UV laser direct writing are combined for fabrication, while finite-difference time-domain simulations guide the microring laser design. The microlasers operate in air and water with thresholds of 0.224 uJ and 0.458 uJ, respectively. Refractometric measurements in aqueous glucose solutions yield a bulk refractive index sensitivity of 364.95 nm/RIU and a detection limit of 2.47 x 10^(-4) RIU. Water-induced polymer changes and photobleaching are also investigated. This fiber-tip architecture provides a basis for probe-based refractometric sensing in environmental and biochemical applications.
Read the full abstract

All authors Mobin Yaghoobi, Seyedeh Mahshad Hosseini, Leila Hajshahvaladi, Gholam-Mohammad Parsanasab, Mohammad Ismail Zibaii

Information Retrieval

Generative Search and Recommendation at Shopify

Search and recommendation at Shopify require retrieving relevant products from a changing catalog of billions of products under strict latency constraints. We present SID-LLM, which casts both tasks as generating semantic IDs: short sequences of content-derived tokens representing products. We jointly pre-train SID-LLM on catalog and behavioral data, giving search and recommendation a shared representation and training objective. We show that reinforcement learning further improves judged search relevance. Furthermore, distillation from task-specific experts improves recommendation quality over joint supervised fine-tuning without a statistically detected loss in search quality. Finally, for efficient serving, we pair multi-token prediction with a lightweight ranker, improving retrieval quality over our autoregressive baseline while reducing latency. We evaluate SID-LLM both as a standalone retriever and as a component of Shopify's production search and recommendation systems. As a standalone search system, SID-LLM achieves quality comparable to Shopify's existing multi-model search stack while meeting production latency requirements. Integrating SID-LLM into the search and recommendation stacks provides significant gains in both offline evaluations and online A/B tests. Together, these results demonstrate the practical value of semantic-ID generation as a unified approach to product discovery at scale.
Read the full abstract

All authors Juhana Kangaspunta, Khalil Hajji, Darshan Fofadiya, Kyumin Park, Michael Garner, Anthony Maltsev, Ryan Synk, Jose Sepulveda, Joshua Crisologo, Shuying Sun, Akshay Soni

Systems and Control

Tags Helping Tags: Coordination-Free Power Boosting for Battery-less Backscatter Networks

Battery-less backscatter sensor nodes powered by radio-frequency wireless power transfer are attractive for large or hard-to-reach deployments, but their range is typically limited by the forward power-transfer link rather than by communication sensitivity. We present a method for improving power delivery, and thus range, in dense tag deployments by having nearby tags act as helpers: each helper terminates its antenna in purely reactive impedances, reflecting the power beacon's signal toward a target tag so that it combines constructively with the direct line-of-sight path and increases the power the target harvests. On a hardware testbed, helpers using discrete impedance control are shown to increase the target tag's harvested power by over 50% above an all-absorbing baseline, while randomly set helper impedances give no reliable benefit, and are sometimes even detrimental. We further propose and implement a local algorithm that lets each helper tune its own impedance using only measurements of the target's packet rate, without channel estimation or coordination between helpers, and we show that it brings the target's packet rate close to that of an independently tuned configuration. These results suggest that independent, passive sensor nodes can cooperatively extend each other's power-transfer range with minimal added infrastructure.
Read the full abstract

All authors Mathew Salas, Xin Xiong, Sihua Shao, Atef Elsherbeni, Mostafa Hassanalian, Rafael Fierro

Computer Vision

Scale as a Modality: Multi-Scale Pixel-Space Diffusion Transformers

Latent diffusion models rely on pretrained autoencoders that introduce lossy compression and constrain spatial granularity. Pixel-space diffusion avoids this bottleneck and enables flexible tokenization at multiple spatial scales, yet existing approaches largely adopt two-level coarse-to-fine architectures with limited cross-scale interaction. We introduce MS-pDiT, a multi-scale pixel-space diffusion transformer that treats spatial scale as a modality. MS-pDiT tokenizes the same noisy image at multiple patch sizes, processes each scale in a dedicated transformer stream, and enables bidirectional information exchange through compacted MM-DiT-style joint attention, with attention cost independent of the finest patch size. This design establishes the number of scales as a new architectural scaling dimension. On ImageNet $256\times256$, increasing the number of scales from two to five consistently improves FID from 4.39 to 2.31, 2.26, and 2.19, while adding an intermediate scale substantially outperforms increasing depth under comparable parameter budgets. Further analysis highlights the importance of bidirectional cross-scale interaction and late introduction of fine-grained streams. MS-pDiT achieves FID scores of 1.43 and 1.44 on ImageNet $256\times256$ and $512\times512$, respectively, and naturally extends to text-to-image generation, achieving 0.86 on GenEval and 83.9 on DPG-Bench.
Read the full abstract

All authors Ruibo Chen, Kaishen Wang, Tianyi Xiong, Chenxi Liu, Yanshuo Chen, Heng Huang, Jian Du

Numerical Analysis

Proof of the Cordero-Torregrosa Conjecture for Methods with Arbitrary Rules and Its Refinement under Paired Function and Jacobian Evaluations

We study the Cordero–Torregrosa conjecture on the order that can be guaranteed when solving nonlinear systems with prescribed budgets of function and Jacobian evaluations. We consider deterministic, causal, memoryless methods whose rules may be discontinuous, nonmeasurable, or noncomputable. We give an independent proof of a scalar barrier for queries of $f$ and $f'$ by constructing a fixed entire function, a simple root, and a single sequence of initial points that simultaneously exclude every exponent greater than $2^{d-1}$, where $d$ is the maximum number of queries. An exact simulation of vector-valued responses yields the Cordero–Torregrosa inequality $p\le 2^{N_F+N_J-1}$ for function classes containing the analytic test fields. The proof holds for $N_F,N_J\ge 0$, $N_F+N_J\ge 1$, without assuming $N_J\le N_F$; in particular, it covers $N_J>N_F$. Under pairing at the end of the cycle, we obtain the sharper bound $p\le 2^{N_F-s+1}3^{s-1}$, $s=\min\{N_F,N_J\}\ge 1$, without requiring temporal precedence. The proofs use indistinguishable perturbations, kernels of constraints, and a construction that preserves responses while producing a single entire witness. These results are upper bounds on the universal local order of a cycle; they do not establish attainability for every budget.
Read the full abstract

Machine Learning

On the Few-Shot Adaptation of Vision-Language Models: A Few Gradient Signs Are All You Need

The limited supervision available for the few-shot adaptation of vision-language models makes gradient estimation particularly noisy. We find that the gradient magnitudes are severely miscalibrated, while the signs of a small top-magnitude subset remain informative. This motivates SignSparseSGD and DynSignSparseSGD, two state-free optimizers that use gradient magnitudes only to select where to update and fixed-magnitude signs to decide how to update. The static variant fixes the support after the first minibatch. The dynamic variant refreshes it at every step. Across 11 datasets, three CLIP backbones, and five shot levels, the proposed methods are competitive with or outperform parameter-efficient and optimization-based baselines on average, while exhibiting markedly lower sensitivity to long training schedules. They introduce no task-specific modules, maintain no first- or second-moment optimizer states, and use one shared configuration across datasets, backbones, and shot levels. Our code is included in the supplementary materials and will be released upon publication.
Read the full abstract

All authors Nairouz Mrabah, Ismail Ben Ayed, Mahdi S. Hosseini, Eric Granger

Artificial Intelligence

The Matthews Correlation Coefficient: Fifty years

Fifty years after its introduction, the Matthews Correlation Coefficient (MCC) remains central -and debated- in binary classification. This survey offers a comprehensive retrospective and an in-depth analysis of MCC, covering its historical development, mathematical foundations, and implications for practice. We clarify long-standing misconceptions, formalize its equivalences (to Pearson's correlation and to Chi2 in 2x2 tables), and examine its behavior under class imbalance. Beyond MCC, we develop analytical (derivatives) and graphical (contour plots) frameworks that generalize to any confusion-matrix-based metric. These frameworks expose structural properties -bias, robustness, sensitivity- that scalar summaries conceal, and they provide actionable guidance for selecting and interpreting performance measures in applications where reliability and fairness matter. Together, our results position MCC within a broader toolkit while clarifying when it is informative -and when alternatives may be preferable.
Read the full abstract

Human-Computer Interaction

Driver Takeovers and Longitudinal Responses Analysis from Naturalistic Driving

As Level 2 advanced driver-assistance systems (ADAS) become more widely used, understanding driver-system interaction is increasingly important for traffic safety. However, limited real-world data make these interactions difficult to study. We present ADAS-TO, a naturalistic dataset of 16,446 takeover clips from 364 drivers across 2,585 different drives. It combines road video, vehicle kinematics, driver inputs, and ADAS status. To analyze takeover behavior, we retained 6,206 manually reviewed takeover events during continued lane keeping or car following. We excluded events in which the driver intentionally changed lanes, turned, or made a stop beyond the ADAS's capabilities (e.g., at a red light or stop sign) shortly after takeover. Drivers began intervening before ADAS disengaged in 65.6% of takeovers involving braking or steering. When the ego vehicle was still closing on a lead vehicle at disengagement, longitudinal safety margins often narrowed further after the transfer to manual control. Among 2,031 such events, 78.5% had a lower minimum time-to-collision (TTC) in the five seconds after disengagement than in the five seconds before it. For events involving braking, the mean TTC trajectory reached its minimum about 1.2 s after disengagement. We model the subsequent braking response using a Gazis-Herman-Rothery (GHR) car-following law. Its median five-second spacing root-mean-square error is 1.34 m with observed lead motion and 1.84 m with predicted lead motion. The fitted model expresses braking as a response to looming, with stronger responses at higher speeds and shorter gaps. Applied before takeover, it shows that greater looming than predicted by the human-response reference is associated with disengagement. These findings support continued assessment of longitudinal safety after ADAS disengagement, while the driver's braking response develops.
Read the full abstract

Machine Learning

MotionBench-XAI: Benchmarking What Shapes Shapley Explanations in Spatiotemporal Machine Learning

Predicting outcomes from spatiotemporal data, such as assessing Parkinson's severity from human motion, often requires explainability in addition to accuracy. Practitioners need to know what drove the prediction, whether it be spatial (a body joint), temporal (a specific window), or both. Shapley values are attributions computed by treating input features as players in a cooperative game and measuring how the model output changes as coalitions of players are removed. Computing them involves choosing what counts as players, how absent players are imputed, and how the game is evaluated. How these choices affect explanation quality is poorly understood for spatiotemporal data, and real data carries no ground-truth attributions to check against. We present MotionBench-XAI, a benchmark built on i) five synthetic spatiotemporal datasets with exact ground-truth Shapley values, ii) three real datasets across clinical gait, electrocardiogram, and audio, iii) ten attribution methods across three player sets, and iv) three trained architectures. We show that grouping features into players, across time, space, or both, matters more than the choice of imputer. Generative imputers improve explanations only when their conditional models are accurate. On real data, the standard proxy metrics select contradictory best methods. We distill these findings into practical insights and provide an open-source pipeline.
Read the full abstract

Mathematical Physics

Constructing Casimirs for the quantum reflection equation

We explore the quantum reflection equation in the case of a full-dimensional $n\times n$ matrix $\mathbb A$. Using the technique of $q$-determinants, we find closed expressions for the quantum Pfaffian (in the case of even $n$) and for new quantum Casimirs that are ratios of special quantum (Sklyanin) minors of $\mathbb A$. Special attention is paid to the self-adjointness property of the newly constructed Casimirs. We put forward a conjecture that, adopting a natural ansatz for a quantum deformation of a Casimir of a classical algebra, self-adjointness implies commutativity. We also explore the relationship to networks of quantum cluster variables.
Read the full abstract

Human-Computer Interaction

MotionDecipher: LLM-Assisted Clinical Interpretation of Human Motion Signals for Efficient Gait Analysis

Clinical gait and motion analysis is widely used to guide targeted interventions for patients with neurological disorders, the outcome of which is typically a report with descriptions of joint movements and forces. Current practices rely on experts' manual effort, which is labor-intensive and time-consuming. Recent advancements in Large Language Models (LLMs) provide opportunities to address this problem, but translating human motion into clinical gait descriptions remains challenging because the quality of the context directly affects LLMs' reasoning capability over continuous motion signals. To tackle this, we introduce MotionDecipher, a novel framework bridging the semantic gap between motion data and clinical interpretations. Specifically, we made 3 key advancements: 1) a Human Motion Tokenizer that converts motion signals into domain-specific tokens for time- and frequency-informed gait interpretation; 2) a Meta Feature Selector that transforms tokens into Structured Representations; and 3) a Hybrid Context Constructor to determine the optimal demonstration-based prompt for In-Context Learning. We evaluated MotionDecipher on data collected at an accredited gait lab from 122 patients and 46 controls. Results show that MotionDecipher achieves 0.91 BERTScore (13% over GPT-5), while expert evaluation ranks MotionDecipher highest in Correctness, Completeness, and Conciseness compared to baselines, highlighting its promise for efficient gait analysis.
Read the full abstract

All authors Mingrui Ma, Tian Tan, Jinsi Guo, Jessica Rose, Yiwen Dong

Machine Learning (Statistics)

High-Dimensional Off-Policy Evaluation without Sample Splitting using an Onsager Correction

Reward models used for off-policy evaluation (OPE) are commonly selected using held-out data. When sample size and context dimension grow proportionally, the difference in test risk between models fitted on the full sample and on a fixed fraction of it need not vanish. Held-out evaluation can therefore consistently estimate the test risk of the model trained on the smaller sample, but because the relative ranking of candidate reward models can change after full-data refitting, held-out selection can be suboptimal. We develop an Onsager-corrected, target-policy-weighted criterion that estimates the full-data reward model's out-of-sample risk under the target policy using only its training observations. This enables early stopping along the full-data gradient-descent trajectory without sample splitting. For neural networks with fixed depth and hidden width under Gaussian contexts, single-index rewards, context-dependent multi-index target policies, and known context-independent logging, we prove uniform consistency of our criterion over a fixed finite set of gradient-descent iterates and a near-oracle guarantee for the selected reward model. The same correction yields a doubly robust estimator that consistently estimates the true policy value without sample splitting, even when reward-model prediction error does not vanish. Experiments show decreasing risk-estimation error as dimension grows at a fixed sample-to-dimension ratio, and improved reward-model selection relative to held-out baselines. Empirically, our correction also reduces the root mean squared error of doubly robust value estimates computed from the same data used to fit and select the reward models.
Read the full abstract

Statistics Theory

Exact optimality of augmented order-of-addition designs

We prove that, for the pairwise-ordering model with at least three components, adding any run to an order-of-addition orthogonal array gives a $D$-optimal design among all exact designs of the enlarged size, allowing replication. We also establish the corresponding result for foldover designs: adding any reversal pair to a foldover order-of-addition orthogonal array gives a $D$-optimal design within the foldover class at the enlarged size. Both augmentations also minimise the average prediction variance and a covariance-scaled information discrepancy, and maximise the minimum relative efficiency. The key step represents the scaled discrepancy through signed comparisons of triples and derives a sharp integer lower bound; classical spectral inequalities then give the exact-size optima, and signed four-component relative-order counts characterise equality. Complete reversal pairs additionally protect the estimation of pairwise-ordering effects against arbitrary reversal-invariant nuisance means. At the smallest estimable five-component foldover size, exact enumeration identifies five determinant-optimal classes with different prediction and balance properties, and a seven-component example shows that deleting a reversal pair need not be optimal even when it minimises the discrepancy. Constructions through ten components illustrate the attained run sizes.
Read the full abstract

Data Structures and Algorithms

Online Bin Packing with Dispatch Deadlines

We study a generalization of online bin packing by introducing dispatch deadlines. Items arrive over time in an online fashion. Each item has an arrival time, a size in $(0,1]$, and a deadline, all of which are revealed only when the item arrives. The algorithm must immediately and irrevocably assign the item to a unit-capacity bin that has not yet been dispatched. Each bin must be dispatched no later than the earliest deadline of the items assigned to it, and once it is dispatched, it cannot be used for future items. The goal is to minimize the number of bins used. Applications arise naturally in logistics. Additionally, the offline version of this problem has appeared under several names, including the clique partition problem in weighted interval graphs and the BinSchedule problem. This problem is also a special case of online bin packing with conflicts and is closely related to online bin packing models with delay costs. We study the competitive ratios of several classical online bin packing algorithms when applied to this problem, and show that a similar hierarchy of competitive ratios carries over. Our main technical results are that the competitive ratio of First-Fit is exactly $16/7 \approx 2.28571$, and that no deterministic online algorithm has a competitive ratio less than $11/5 = 2.2$, showing that First-Fit is quite close to best possible. In addition, Next-Fit and Worst-Fit have competitive ratio exactly $3$, and every algorithm in the Any-Fit class has competitive ratio between $16/7$ and $3$.
Read the full abstract

All authors Pei-Lin Chiang, Wing-Kai Hon, Kazuo Iwama, Li-Tzu Yang

Sound

AircraftWitness: How Often Aircraft Can Locate Recordings

Aircraft overhead can reveal where an outdoor recording was made. Propellers and rotors leave Doppler curves in audio, and public ADS-B archives log equipped aircraft. We measure how often, blind and against chance. Our main test was registered and timestamped before any audio, and its two amendments came before any answer was read. It covers 24 new places, with the recorders' own clocks, a 300 s slack and a 20 km square. The best of three locators places clear-pass clips within 1 km at 4 places, where a uniform guess predicts 0.57 (p = 0.0065 after correcting for the pick). With the wrong time or swapped audio it places none. The other two hit 1 and 2. On 360 clips from 7 places of a GPS-timed network, two locators place 6% and 8% within 1 km. By place, a unit chosen after that run, they hit 4 and 5 of 7 where a uniform guess predicts 2.3. On 1,600 random public recordings, picked and cut without the truth in a time-bounded run, the three locators answer at most 20 times and place at most 2. We release AircraftWitness, a benchmark, and pass labels for 11.7 million recordings.
Read the full abstract