Machine Learning
Reachability-Based Formal Verification of Graph Neural Networks with Node and Edge Features
Anne M. Tumlin, Ben Wooding, Zhenxuan Shao et al. (+3 more)
Graph neural networks (GNNs) have become a prominent approach for developing fast,
topology-aware surrogates in electric power systems, supporting tasks such as power flow (PF) analysis, optimal power flow (OPF) estimation, and cascading failure analysis (CFA). Despite this growing use, formally verifying GNN-based models remains challenging, with existing methods limited in scope. We extend the neural network verification (NNV) framework to graph-structured inputs through GraphStar sets, a generalization of
Star sets that captures uncertainty over both node and edge features. This extension enables the propagation of linear message-passing operations and the sound approximation of ReLU nonlinearities for GNN architectures, including graph convolutional network (GCN) and graph isomorphism network with edge features (GINE) layers. We evaluate GNNV across three power system tasks, PF, OPF, and CFA, on the IEEE-24, IEEE-39, and IEEE-118 test cases, as well as two standard graph classification
benchmarks, ENZYMES and
PROTEINS. Our results show that GNNV provides tighter robustness guarantees than CORA on graph classification models with ReLU-based activations and, for the first time, delivers edge-aware robustness guarantees for GINE-based PF and OPF models under joint node and edge perturbations.
Read the full abstract ▾
All authors Anne M. Tumlin, Ben Wooding, Zhenxuan Shao, Diego Manzanas Lopez, Tyler Derr, Taylor T. Johnson
Machine Learning (Statistics)
Calvin Tolbert
Nuclear-norm-regularized least squares is a standard approach to
matrix completion, but it yields only a point estimate. Its
Bayesian counterpart, a prior combining the
nuclear norm with a Frobenius-norm term, gives a strongly log-concave though nonsmooth posterior when the noise precision is known. In practice the precision is unknown, and placing a truncated Jeffreys prior on it makes the joint posterior neither log-concave nor smooth. To our knowledge, we give the first sampler for this model with an explicit non-asymptotic guarantee: its output is within a prescribed total-variation or Wasserstein distance of the posterior on the
matrix after a number of steps
polynomial in the
matrix dimensions and the reciprocal of the target accuracy. The sampler never targets a continuous distribution that is not strongly log-concave. It discretizes the precision onto a grid, estimates its posterior by thermodynamic integration along the tilted conditional posteriors of the
matrix, and bounds the error this propagates to the posterior on the
matrix. The approach applies whenever non-log-concavity is confined to a bounded scalar hyperparameter and the conditional distributions are strongly log-concave with smooth-plus-Lipschitz potentials. Our result establishes feasibility; the
polynomial complexity is not intended as a practical
algorithm at current problem scales.
Read the full abstract ▾
Image and Video Processing
Rate-distortion optimization for full-reference image quality metrics via stochastic Hessian estimates
Samuel Fernández-Menduiña, Eduardo Pavez and Antonio Ortega
Block-based
video codecs select coding parameters based on the input by optimizing a rate-distortion trade-off. The conventional distortion choice, the sum of squared errors (SSE), simplifies parameter selection: the SSE is the sum of block-wise SSEs, so rate-distortion optimization (RDO) can treat blocks independently. Alternatively, full-reference image quality assessment (FR-IQA) metrics such as MS-SSIM or LPIPS often align better with the human visual system than SSE, but they cannot be used in-loop: they do not decompose block-wise and typically require the fully decoded image as input. Building on existing results in metric quadratization, we approximate a broad class of FR-IQA metrics by an input-dependent quadratic distortion (IDQD), whose quadratic form
matrix is derived from the Hessian of the metric evaluated at the source
video. To make the distortion computable block-wise, we propose two approximations of the Hessian
matrix: 1) keeping the block-diagonal, and 2) keeping only its diagonal. We propose
estimators for both that require only
matrix-vector products with the Hessian obtained by automatic differentiation. Across five metrics for Kodak and CLIC in VVC, IDQD-RDO achieves 14.2-36.7 % BD-rate savings under the target metric with no decoder changes and incurs 10-30 % encoding complexity overhead.
Read the full abstract ▾
Human-Computer Interaction
Beyond Driving: Envisioning Activities in Future Autonomous Vehicles through Experience-Centered Design
Keqi Chen, Xiao Xue, Xinyi Liu et al. (+4 more)
Autonomous vehicles (AVs) are poised to fundamentally alter personal transportation, offering occupants the freedom to engage in various non-driving-related activities (NDRAs). However, our current understanding of how people might actually use this time in fully autonomous vehicles (FAVs) is limited. Traditional research methods often struggle to capture the influence of diverse travel circumstances and purposes when exploring future scenarios. This paper introduces experience-centered design (ECD) as an approach to investigate potential NDRAs within FAVs by examining the intricate connections between individuals' daily routines, specific travel contexts, and the activities they might undertake in transit. Through a multi-phase study employing participatory techniques that facilitated narrative construction and exploration, including diary studies, scenario scripting, and mixed reality (MR) enactments, we enabled participants to ground speculative future scenarios in their own lived experiences. This process yielded nuanced insights into preferences and behaviors concerning potential NDRAs, alongside the underlying subjective meanings and sociotechnical considerations. Our findings lead us to conceptualize NDRAs not as isolated instances of "travel time use," but as dynamic sequences of interrelated activities deeply shaped by pre- and post-journey contexts. The effectiveness of our ECD approach in bridging current lived experiences with future scenarios was crucial for uncovering these insights. Ultimately, this study reconceptualizes AVs as complex sociotechnical systems that actively mediate human activity and interaction, suggesting a fundamental shift in their role within the urban fabric.
Read the full abstract ▾
All authors Keqi Chen, Xiao Xue, Xinyi Liu, Runjia Tan, Chris Speed, Lee Kwan Min, Chen Lv
Computation and Language
A Native-Reference Phone-Class Geometry for Second-Language Pronunciation Analysis
Tina Raissi, Nhan Phan, Chenxiao Wang et al. (+1 more)
Automatic speaking assessment systems can provide holistic proficiency scores, but often lack interpretable measures that characterize pronunciation quality. We propose a native-reference phone-class geometry for measuring second
language (L2) pronunciation deviation without requiring pronunciation labels, read-aloud
prompts, or matched recordings of the same text from native and L2 speakers. Given a native
speech corpus, we average frame-level self-supervised representations for each context-dependent phone-class and use singular value decomposition (SVD) to derive a compact native-reference coordinate system. For each L2 utterance, we compute the corresponding averages and project them into the native-reference space. We then demonstrate that the distances between L2 and native-reference coordinates for matched phone-classes show consistent negative correlations with holistic speaking proficiency on the Dev subset of the Speak and Improve Corpus 2025 (Spearman's $\rho\!=\!-0.53$) and with pronunciation quality on the learner subset of the English Read by Japanese Students
dataset ($\rho\!=\!-0.34$). These findings suggest that the proposed geometry captures acoustic-phonetic information relevant for proficiency rating while remaining applicable to spontaneous L2
speech without matched native recordings.
Read the full abstract ▾
All authors Tina Raissi, Nhan Phan, Chenxiao Wang, Mikko Kurimo
Computation and Language
How Reproducible Are Evaluation Conclusions? A Self-Audit of LLM-Inferred Prompt Structure
Dipankar Sarkar
Evaluations of LLM systems routinely average over small
prompt sets and report models as a ranked table. We ask how much confidence such a table deserves, using LLM-based
prompt-structure inference as the case study: eight open model variants across five families and 8B to 675B parameters, caching disabled, 293 raw intermediate representations persisted. The measured phenomenon is unstable to begin with. Identical calls do not reliably recover identical structure, with mean node-set Jaccard from 0.39 to 0.96 and 72% of
prompt-model
cells never node-set-perfect.
Auditing the evaluation weakens its conclusions further, and this is our main contribution. Under a joint cluster bootstrap over
prompts, only the bottom of the ranking is firm: the two least reproducible models hold rank in 99% and 86% of replicates, the middle four in 27% to 48%, and the top two in 68% each, so the table identifies the worst model reliably but does not reliably identify the best. Two equally defensible rules for merging repeated campaigns change four of eight rows and move the study-wide headline by 7 percentage points. Checking the inferred structure against ground-truth annotations shows reproducibility cannot be read as accuracy. And four of the eight endpoints were withdrawn within ten weeks of measurement, so the study as specified can no longer be run. Small-sample LLM evaluations can therefore look far more definitive than their evidence supports. We recommend reporting rank stability, per-
cell provenance, executed sensitivity comparisons, raw per-run outputs, and a measurement date alongside any ranking.
Read the full abstract ▾
Quantum Algebra
The crystal of the categorified quantum group
Soo-Hong Lee
A categorification of the
crystal $B(\widetilde{U}_q(\mathfrak{g}))$ of the modified quantized enveloping
algebra associated to a symmetrizable Kac-Moody
algebra $\mathfrak{g}$ is given. A
crystal is constructed from a categorical $\mathfrak{g}$-action on abelian categories without finiteness conditions, and is used to construct the bicrystal structure on the set of self-dual indecomposable objects of the categorified
quantum group.
Crystals from Webster's generalized tensor product categorification are identified with tensor products of lowest and highest weight
crystals. As an application, a categorification of the extremal weight module is obtained.
Read the full abstract ▾
Statistical Methodology
PK/PD-integrated Bayesian platform design for phase II dose regimen optimization
Axel Vuorinen, Antoine Guillon, Emmanuelle Comets et al. (+1 more)
Early-phase dose-finding methods increasingly assess toxicity and efficacy jointly, but comparisons based only on administered dose may inadequately characterize regimens differing in schedule. We developed a
Bayesian phase II adaptive platform design for regimen optimization that integrates pharmacokinetic/pharmacodynamic (PK/PD) modelling into toxicity, efficacy, regimen selection and adaptation decisions. The proposed PK/PD-informed Regimen Optimization Platform (PROP) design uses a population PK/PD model to generate patient- and population-level predictions of exposure and biological activity. Acute and cumulative toxicities are analysed using a discrete-time time-to-event model informed by PK exposure. Efficacy is evaluated through
Bayesian model averaging of exposure-driven and biomarker-driven time-to-event models. The design supports regimen graduation, discontinuation for futility or safety, and addition of unexplored regimens. Performance was evaluated through simulations motivated by an influenza intensive-care setting. Across six scenarios, PROP generally improved graduation and futility decisions, reduced inappropriate graduation, and supported the addition of promising regimens compared with dose-based alternatives. It also more accurately estimated regimen-specific toxicity and arm-specific efficacy, while the model-averaging framework favored the efficacy model consistent with the data-generating mechanism. Dose-based approaches performed better for safety stopping in some scenarios, despite less accurate characterization of the regimen–toxicity relationship. PK/PD-informed platform designs can improve adaptive regimen selection and knowledge generation when dose alone cannot adequately characterize treatment regimens.
Read the full abstract ▾
All authors Axel Vuorinen, Antoine Guillon, Emmanuelle Comets, Moreno Ursino
Computation and Language
Scoring Both Directions: LLMs realize the MRS they cannot reliably parse
Soham Dan
The English Resource Grammar (ERG) is a hand-written computational grammar of English. Given a sentence, its processor, ACE, produces a formal meaning representation called Minimal Recursion Semantics (MRS): a graph of the sentence's predicates and their arguments. The grammar is bidirectional and can also turn an MRS back into an English sentence. \citet{hajdik2019} used the ERG's treebank to build a
benchmark for that generation task, MRS to text, and trained sequence-to-sequence models to solve it. The parsing task, text to MRS, can be tested on the same sentences. We reconstruct their 10K-sentence test split, and score two large
language models, Claude Sonnet~4.5 and Claude Opus~5, in both directions against their trained systems and against ACE, with no task-specific training. Given an MRS and three examples, Opus writes the sentence at 76.3 BLEU, ten points above their system trained on 72k pairs (66.1 BLEU), and comparable to their system trained on a million extra pairs (77.2 BLEU). Sonnet scores 65.7 BLEU, and letting it choose among ACE's own candidate sentences lifts it to 69.6, while a pooled judge that keeps Opus's own sentence among the candidates adds 0.6 points (77.0 BLEU). In the parsing direction, however, the models fall far behind ACE: asked for the MRS of the same sentences, they reach 57.2 (Sonnet) and 65.5 (Opus) F$_1$ on the graph's predicates and arguments against 91.0 for ACE, and exact-match the gold on about 1% of sentences. We characterize the failure modes for the parsing tasks, and conclude that a generation score alone does not show that models understand formal semantic representations.
Read the full abstract ▾
Cryptography and Security
A Data-Driven Analysis of Infostealer Malware Victims
Arttu Paju, Juha Nurmi, David Arroyo et al. (+7 more)
Infostealer malware infects devices worldwide and harvests their most sensitive contents: credentials, browser sessions, private keys, and access certificates. Yet its impact on victims remains difficult to study without an ethical, legal, and curated research
dataset. To close this gap, we build a
privacy-preserving pipeline that turns illicitly sourced infostealer logs into a reproducible research artifact, minimizing sensitive data while preserving measurement utility, and use it to construct a
dataset of 170,298 victims from logs of multiple infostealer families. Analyzing these victims, we find that the most compromised services mirror the world's most popular platforms, with gaming and entertainment services strongly overrepresented. Within the sample we identify compromised credentials for high-value organizations, including law-enforcement domains, government and military services, and all eight Ivy League universities, as well as substantial exposure of security-critical infrastructure and of financial, remote-access, and development platforms. Victims also show widespread credential reuse and significant revictimization risk, overlapping with phishing and ransomware victim populations. We release the first anonymized victim-level infostealer
dataset under controlled access to enable ethical,
privacy-preserving, and reproducible research on information security and victim behavior.
Read the full abstract ▾
All authors Arttu Paju, Juha Nurmi, David Arroyo, Sergio Chica Manjarrez, Fran Casino, Mikko Niemelä, Juuso Itkonen, Joel Scanlan, Constantinos Patsakis, Georgios Smaragdakis
Quantum Physics
Design Principles for Ultra-High-Rate Quantum Codes
Jong Yeon Lee, Koki Okada, Nishad Maskara et al. (+2 more)
Reducing the
qubit overhead of
quantum error correction is a central challenge for scalable fault-tolerant
quantum computing. Recent ultra-high-rate
quantum codes offer a promising route toward this goal, with some constructions requiring as few as two physical data
qubits per logical
qubit. However, systematic principles for navigating the tradeoffs among encoding rate, distance, check weight, and blocklength remain lacking. Here, we develop and analyze principles for exploring this design space, and use them to design compact code constructions with improved performance. We develop code templates based on a pair-partition construction and a halving transformation that further reduces blocklength. Motivated by ensemble analysis of the degree distributions, we identify column weight as a key design parameter: increasing the column weight enables larger distances at compact blocklengths, at the cost of heavier checks. We find that at physical error rates of 0.1%, the benefits of increased distance often outweigh the penalty associated with heavier checks. Applying this framework, we identify numerous compact codes with favorable parameters, including [[90,21,11]], [[140,31,15]], and [[200,43,20]] non-CSS codes with check weight 10. Moreover, we develop
symmetry-informed strategies for identifying low-weight logical bases. These results provide systematic strategies for designing ultra-high-rate
quantum codes and navigating their Pareto frontier.
Read the full abstract ▾
All authors Jong Yeon Lee, Koki Okada, Nishad Maskara, Kenta Kasai, Hengyun Zhou
Combinatorics
S-meandric Permutations and Tangency Polynomials
Yury Belousov and Viktoriia Georgievskaia
A meander is a configuration of two simple plane curves intersecting transversely. The orders of their intersection points define a permutation that determines the configuration. When tangencies are allowed, however, different configurations can share the same permutation. We study the combinatorial and algebraic structures arising from this non-uniqueness. We give a realization criterion and show that the realizations of each realizable permutation form an affine space over the two-element field. We describe this space using an associated graph, called the component spine. We prove that the component spine of every permutation is a cactus. We also introduce the tangency
polynomial, which counts realizations by their number of tangencies, investigate its properties, and prove that it factors over the cycles and bridges of the component spine. We derive a central limit
theorem for tangency counts and obtain asymptotic formulas for the number of distinct tangency
polynomials.
Read the full abstract ▾
Dynamical Systems
Hyperbolic Area Methods in Meromorphic Dynamics and Elliptic Polynomial Skew Products
Zihao Ye
We use
hyperbolic area and covering geometry to study wandering Fatou components and singular orbits. We give proofs without quasiconformal deformation of the rational no-wandering
theorem and the known no-wandering results in Bergweiler's Question~9. The framework also excludes wandering for entire functions with compact singular sets whose derived sets lie in the Fatou set. For Bergweiler's Question~8, every wandering component of a transcendental meromorphic map has a subsequence of iterates converging locally uniformly to infinity. For Bergweiler's Question~4, we establish
hyperbolic separation near singular-free Baker cycles after deleting any finite set of original singular values as orbit generators. For Bergweiler's Question~11, corresponding finite-deletion boundary estimates hold along wandering tails. In higher dimension, a relative-area
theorem excludes bounded wandering components of
polynomial skew products over bounded bases carrying an absolutely continuous invariant probability of full support. Applied to
elliptic polynomial skew products over Siegel rotation domains, it yields a local no-wandering
theorem without conditions on the fiber critical orbits or continuity of fiber Julia sets. A global criterion covers regular skew products on $\mathbb P^2$ whose bounded base Fatou components eventually enter periodic Siegel disks, including $(\lambda z+z^d,q(z,w))$ for Brjuno $\lambda$ and
polynomial $q$ of total degree at most $d$ with a nonzero $w^d$ coefficient.
Read the full abstract ▾
Quantum Physics
Beyond Bond Gauge: Exact Tensor-Network Tangent Spaces at Weighted Graph States
Song Cheng
For normal PEPS, the fundamental
theorem settles at the finite level whether virtual-bond gauge exhausts a tensor network's representation freedom. We ask the first-order question at a fixed representation, from both sides of the differential: which tensor variations leave the state unchanged without being bond gauges, and which physical directions can no variation reach? We answer both exactly for weighted graph state (WGS) tensor networks with equal physical and bond dimension $q\ge2$, which include the graph-state resources of measurement-based
quantum computation. Dividing a single-tensor variation by the nonvanishing amplitude turns it into an arbitrary function on the closed neighborhood of its vertex. Overlapping neighborhoods give the complete kernel. Its non-bond part is generated on periodic regular-polygon tilings by triangles, chordless squares, and edge-sharing diamonds. Under stated period conditions, generic gauge completeness on all thirty-nine families studied identifies this nullity as a WGS rank drop: for thirty-three families at every $q\ge2$, and for six at $q=2$. Supports outside every neighborhood give the missing directions: at
qubit graph states on the eleven Archimedean tilings every nearest-neighbor $XX$ and $YY$ direction is missing, including on honeycomb, where the non-bond kernel is empty. On large honeycomb and square
tori, small nonzero coherent $XX$ errors leave the bond-two PEPS set on the same graph. Removing redundant parameters and supplying missing directions are therefore distinct operations.
Read the full abstract ▾
Exactly Solvable and Integrable Systems
Superintegrability of discrete-time rational Ruijsenaars-Schneider model and deformed polynomial symmetry algebras
Pavel Drozdov
We explicitly construct the additional integrals of motion, ensuring maximal superintegrability of the discrete-time rational Ruijsenaars-Schneider model. Using them, we investigate the algebraic aspects of superintegrability in both continuous- and discrete-time settings. In particular, we determine the complete structures of the
polynomial symmetry algebras associated with both the rational Ruijsenaars-Schneider model and its discretization. We demonstrate that discretization leads to a nontrivial deformation of the continuous
symmetry algebra with respect to the discretization parameter, thereby extending recent analogous results from the rational Calogero-Moser system to its relativistic generalization.
Read the full abstract ▾
Computational Complexity
Sharp Lovasz-Theta Bounds on Random Graphs
Aaron Potechin and Jeff Xu
It is well known that the \Lovasz-Theta function of a random graph $G(n,\tfrac{1}{2})$ is $\Theta(\sqrt{n})$. More precisely, it is tightly concentrated in the interval \( [\sqrt{n},\, 2\sqrt{n}], \) where the upper bound follows from an explicit dual witness for the associated semidefinite program. Numerical evidence and heuristic arguments suggest that the true value is $(1+o(1))\sqrt{n}$. However, closing this gap has remained a longstanding challenge, resisting existing techniques even in light of recent progress on sharp algorithmic thresholds and non-asymptotic free probability. In this work, we resolve this question by proving that the \Lovasz-Theta function of $G(n,\tfrac{1}{2})$ is $(1+o_n(1))\sqrt{n}$ with high probability, determining its asymptotic value up to vanishing relative error.
Read the full abstract ▾
Artificial Intelligence
Self-Play Pretraining with Zero Data
Aditya Cowsik, Kfir Dolev, Michael Y. Li et al. (+4 more)
Advances in
language modeling have been driven by scaling pretraining on ever more data. Yet, the training data is still largely curated on the model's behalf. A more general approach to pretraining would let the model learn to generate the data most useful for its own improvement. This would provide an effectively unbounded source of training data, limited by compute rather than human knowledge. We introduce Self-Play Pretraining with Zero Data, an initial proof-of-concept towards realizing this
vision. Our procedure casts synthetic data generation as a search over the space of all computable structure, taking inspiration from Solomonoff induction. Starting from random initialization, two models learn in tandem: a generator proposes programs interpreted by a universal Turing machine, generating byte sequences, while a learner autoregressively predicts these byte sequences. The learner is trained with standard cross-
entropy, while the generator is trained with reinforcement learning to produce sequences at the frontier of the learner's capabilities, yielding an adaptive curriculum. A universal Turing machine gives us a search space over all computable data-generating processes, imposing little domain-specific structure, and self-play searches over this space for useful training data. We test whether zero-shot performance on natural data improves predictably with self-play compute; this is a clean test of transfer since neither generator nor learner is trained on natural data. Across several natural
datasets, zero-shot loss exhibits predictable scaling in compute. The models also exhibit in-context learning, and discover recognizable mathematical sequences during training.
Read the full abstract ▾
All authors Aditya Cowsik, Kfir Dolev, Michael Y. Li, G. Bruno De Luca, Nourya Cohen, Noah D. Goodman, Yoav Levine
Programming Languages
Lifting the Preprocessor with Oxidize: Structure-Preserving C-to-Rust Translation (Technical Report)
Robbe De Greef, Théo Engels, Felix Van den Broucke et al. (+2 more)
This technical report accompanies our EuroSys 2027 paper "Lifting the Preprocessor with Oxidize: Structure-Preserving C-to-Rust Translation" (DOI 10.1145/3842654.3848533). Oxidize is a C-to-Rust translator that retains C macros as Rust macros instead of expanding them away, so that the translated code keeps the abstractions its developers wrote. The report collects material the paper cites but has no room for: the translator's intermediate representation, pass structure and back-end emission; the details of macro typing through helper traits and the generic cast macro; the sources and pinned revisions of the nine evaluated C codebases; the per-codebase macro population and how macros are attributed to classes; a roadmap from structure-preserving translation toward safe Rust; the performance overhead study and its methodology; and the configuration and
prompt used for the LLM-based baselines.
Read the full abstract ▾
All authors Robbe De Greef, Théo Engels, Felix Van den Broucke, Ken Hasselmann, Antonio Paolillo
Logic
Existence of a Model Companion for Groups of Exponent 3
Yawara Ishida, Ryosuke Mizuno and Kota Takeuchi
In this article, we prove that the theory $T_3$ of groups of exponent $3$ has a model companion. Previous work established the existence of model companions for theories of groups of fixed finite exponent and nilpotency class at most $2$ (Saracino–Wood), and for theories of groups of
prime exponent $p$ and nilpotency class at most $c<p$ (Maier). These results do not cover $T_3$, since groups of exponent $3$ may have nilpotency class $3$. Our
theorem verifies the $n=3$ case of a
conjecture proposed by the third author that, for each integer $n>1$, the theory $T_n$ of groups of exponent $n$ has a model companion if and only if every finitely generated group of exponent $n$ is finite, or equivalently, if and only if the Burnside problem has a positive solution for exponent $n$.
Read the full abstract ▾
Optics
Angular-Channel Capacity and Nonlocal Wavefront Engineering for Phased-Array–Lens Systems
Mohammad Soltani and George V. Eleftheriades
Wide-angle beam steering with high directivity is fundamentally limited by diffraction, restricting the scan performance of phased-array systems. Here, we establish the bounds governing beam steering with high directivity in phased-array–lens systems and show that nonlocal dielectric metalenses can operate close to these limits. Unlike conventional metalenses that suppress mutual coupling, our approach exploits nonlocal electromagnetic interactions to laterally redistribute energy and expand the spatial width of phase coherence across the aperture. This enables efficient wavefront engineering for uniformly excited planar phased arrays using conventional linear progressive phasing. We further reveal the roles of diffraction and evanescent waves in determining the maximum coherent-aperture expansion and directivity enhancement. The resulting angular-channel bounds identify two distinct operating regimes for passive linear phased-array–lens systems: scan-resolution enhancement and directivity enhancement with preserved scan range. Guided by these limits, we derive geometry-independent design equations and develop an adjoint-based inverse-design framework for passive linear metalenses, enabling planar and cylindrical implementations that operate close to the fundamental limits. We experimentally validate a compact cylindrical metalens that preserves the $50^\circ$ scan range and scan resolution of a phased array with a $3.5\lambda_0$-wide aperture while providing a $2.5$–$3.5\,\mathrm{dB}$ increase in peak directivity. These results establish nonlocal dielectric metalenses as a practical platform for diffraction-limited wavefront engineering in phased arrays.
Read the full abstract ▾
Distributed Computing
KernelOPT: Dispatch-Aware Agentic Search for GPU Kernel Optimization
Aheli Poddar, Sanskar Prasad, Arindam Samanta et al. (+3 more)
Deep learning inference and training performance depends critically on GPU kernel efficiency. Modern compilers such as PyTorch Inductor automatically generate GPU kernels from high-level model code, but frequently underperform expert-written implementations by wide margins. Recent LLM-assisted kernel optimizers can close this gap for standalone kernels, yet treat compiled models as black boxes, generally optimizing individual standalone kernels without respecting the compiler's structural decisions or verifying the model end-to-end. We present KernelOPT, a multi-
agent system that treats compiled models as structured artifacts. It preserves vendor library calls (cuBLAS, cuDNN) and exclusively targets generated Triton sub-kernels using five profiling-guided LLM
agents. A four-gate verification cascade applies static validation, multi-seed correctness checking, model-level float64-fallback verification, and performance gating ($\gamma{=}1.03$) to filter candidates and verify the re-stitched model end-to-end. When candidates fail verification, the system preserves the compiler baseline. The system accepts PyTorch nn Modules, standalone Triton kernels, and Helion kernels. Evaluated on 250 KernelBench problems (100 Level 1, 100 Level 2 and 50 Level 3) on NVIDIA H200, KernelOPT achieves
geometric mean speedups over torch compile of 1.40$\times$ (L1), 1.15$\times$ (L2), and 1.07$\times$ (L3) across all kernels, including fallback cases. Optimized-only geomeans (excluding cases where verification gates preserve the compiler baseline) are substantially higher: 2.54$\times$ (L1: 36/100), 1.84$\times$ (L2: 23/100), and 1.37$\times$ (L3: 11/50), reflecting where the optimizer achieves meaningful leverage.
Read the full abstract ▾
All authors Aheli Poddar, Sanskar Prasad, Arindam Samanta, Subha Chakraborty, Vishal Goyal, Rohit Singh Rathaur
Game Theory
Can Labor Markets Function in the Age of AI? The Evaluation Bottleneck in Hiring
Itai Ashlagi, Ramesh Johari, Jon Kleinberg et al. (+1 more)
AI-assisted job-search tools have become increasingly popular by making it easier to find and apply to jobs. But by making it easier for applicants to generate and tailor application materials, they can also reduce how informative those materials are about applicant fit. We study this tradeoff in a hiring
market where applicants differ in experience and latent match quality and firms use noisy application materials to decide whom to screen. We ask how AI affects downstream screening and hiring, and which applicants are most adversely affected. As application materials become less informative, a
Bayesian firm rationally relies more heavily on coarse observables such as prior experience. Among the four applicant types defined by experience and compatibility for the job, inexperienced-compatible applicants are the most exposed: they lack observable experience and lose the individualized information that could distinguish them from other inexperienced candidates. When screening is costly, these changes can also generate inefficient screening failures in which firms screen no applicants or screen only experienced applicants. We then show that multistage hiring can arise as an endogenous firm response: a relatively inexpensive intermediate assessment allows firms to acquire new evidence of fit before costly full screening. This can restore screening opportunities that disappear under one-stage hiring and give inexperienced-compatible applicants a path to screening. Our results show how AI can shift the central friction in hiring from submitting applications to obtaining credible evaluation, creating entry barriers for high-fit workers without prior experience. Multistage hiring can endogenously arise in response, restoring evaluation opportunities that would otherwise disappear and helping preserve
market functioning.
Read the full abstract ▾
All authors Itai Ashlagi, Ramesh Johari, Jon Kleinberg, Anushka Murthy
Distributed Computing
KREX: Concurrent Kernel Benchmarking on Shared GPUs via Region-Granular Exclusivity
Tianyu Feng, Haoxuan Yu, Tianyuan Wu et al. (+8 more)
LLM
agents automate GPU kernel optimization by repeatedly composing candidates and measuring their duration on real GPUs. Existing systems preserve measurement fidelity by reserving a GPU for an entire
agent session or benchmarking command. However, this results in poor utilization because only a small fraction of command execution requires exclusive GPU access. Sharing GPUs could recover this idle capacity, but introduces contention that compromises measurement fidelity and misdirects the
agent's search. We present KREX, a runtime for concurrent kernel
agent benchmarking with region-granular exclusivity. KREX lets
agents mark critical regions involving timing-sensitive operations within a benchmarking command. The runtime then enforces exclusivity within marked regions and allows concurrent execution outside them, achieving high throughput while preserving measurement fidelity. To enforce in-region exclusivity, KREX blocks new competing GPU submissions and drains outstanding work before freezing sibling processes and isolating CPU cores, protecting both GPU execution and the host threads that drive measurements. To maximize off-region concurrency, KREX reuses GPU contexts in persistent context processes to avoid repeated, node-wide serialized context creation. We evaluate KREX on NVIDIA and AMD GPUs. Compared with command-granular exclusivity baselines, KREX delivers up to $3.4\times$ the benchmarking throughput with a negligible p95 timing inflation of $0.30\%$, $1.58\%$, and $3.90\%$ for kernels longer than 10 ms, 1 ms, and 0.1 ms, respectively.
Read the full abstract ▾
All authors Tianyu Feng, Haoxuan Yu, Tianyuan Wu, Lingyun Yang, Daocheng Ying, Yuxiao Wang, Ruibo Fan, Yinghao Yu, Guodong Yang, Liping Zhang, Wei Wang
Robotics
M3GD: Multi-Modal Multi-View Geometric Diffusion for Camera–LiDAR Novel View Synthesis
Yang Zhou, Jiuhong Xiao, Shizhao Ye et al. (+3 more)
Robotic novel view synthesis (NVS) must recover both visual appearance and metric 3D structure, yet most generative NVS methods rely only on images, overlooking
LiDAR, a complementary sensor common on robotic platforms. We present M3GD, a Camera–
LiDAR multimodal representation for generative NVS that composes independently pretrained 2D image and 3D point-cloud foundation models without separately pretraining a cross-modal translator. We show that, after camera projection, frozen
LiDAR and image features exhibit substantial shared spatial structure, providing a natural cross-modal representation. M3GD conditions generation on
LiDAR through this structure: it combines explicit geometry statistics with learned point-cloud descriptors into view-aligned packets on the image-latent grid, injected through a lightweight residual adapter into a multi-view flow-matching generator whose latent space, decoders, and training objective remain intact. On the GrandTour
dataset, M3GD improves target-view RGB and depth synthesis over an image-only version of the same backbone. Ablations show that the gains come from pixel-aligned
LiDAR content and that target-view
LiDAR acts as a
geometric query linking the requested view to source observations. Deployment on a ground
robot demonstrates practical real-world operation, with a configurable quality–cost trade-off controlled by the number of Euler integration steps.
Read the full abstract ▾
All authors Yang Zhou, Jiuhong Xiao, Shizhao Ye, Long Quang, Carlos Nieto-Granda, Giuseppe Loianno
Software Engineering
Era by Eon: Benchmarking Enterprise Agents on Hidden Knowledge
Benjamin Gruenbaum, Doron Porat, Assaf Natanzon et al. (+3 more)
In the Era by Eon
benchmark, each question states the rules for its answer, and code computes the answer from a generated company's data. When
agents can run code, the four strongest models each answer 22 to 25 of 27 such questions, so the
benchmark barely separates them. We add eight question templates that depend on hidden facts. No question or document states a hidden fact, and the records that seem to hold it show something else. Other data implies it. For example, the sales system says a customer dropped a purchase because of timing. On a recorded call, the customer blames an outage. For each generated company, code fills each template and computes an exact answer without a
language model. We evaluate 12
agents. Each pairs a model with an
agent program, which connects it to the company's systems. The best
agent answers 18 of its 24 attempts, three per question, correctly. Four of the six models answer at most 6 of 24 with any program. The hardest questions require picking one of several similar records, such as which of three renewal offers a customer signed. All
agents together answered two such questions correctly in only 1 of 84 attempts.
Read the full abstract ▾
All authors Benjamin Gruenbaum, Doron Porat, Assaf Natanzon, Roy Zavida, Chen Dinachi, Or Itzahary
Artificial Intelligence
SciWalker: Synthesizing Scientific Coding Problems with Operator Graphs and Execution Feedback
Chenxi Li, Wenxuan Zeng, Yun Luo et al. (+4 more)
Improving the scientific coding capabilities of large
language models (LLMs) requires high-quality training data. However, such data remain scarce because manually authoring realistic problems is costly and time-consuming, while systematically covering diverse scientific domains and algorithmic combinations remains challenging. To address this, we introduce SciWalker, a framework for synthesizing scientific coding problems through operator-chain sampling and execution feedback. The framework combines scientific library interfaces with operation modes to instantiate operators, organizes them into operator graphs, and samples operator chains as computational workflow cues. Guided by these cues, we adopt LLMs to generate scientifically grounded problem statements, reference solutions, and tests, with failed generations iteratively repaired using execution feedback. By combining structured workflow composition with verification and quality review, SciWalker enables scalable task generation while promoting scientific grounding, computational diversity, and executability. Using this framework, we construct 8,178 high-quality problems spanning 5 scientific domains and 32 subdomains. To evaluate their training utility, we conduct reinforcement learning on Qwen3.5-9B using the GSPO
algorithm. This training improves SciCode subproblem accuracy by 9.9 percentage points, from 29.3% to 39.2%, with gains across scientific code generation, code repair, and
reasoning benchmarks. The code for SciWalker is available at https://github.com/lichenx1/SciWalker.
Read the full abstract ▾
All authors Chenxi Li, Wenxuan Zeng, Yun Luo, Fangchen Yu, Peng Ye, Yu Cheng, Jun Zhang
High Energy Astrophysical Phenomena
A Helical Trajectory Interpretation of a Multi-wavelength Event in a Blazar S5 0716+714
Zhihao Ouyang, Jingyu Wu, Hubing Xiao et al. (+9 more)
Relativistic jets in blazars exhibit strong variability across multiple timescales, but the connection between temporal variability and the
evolution of jet structures remains poorly understood. In this work, we investigate the $\gamma$-ray variability of the BL Lac object S5 0716+714 from July 2014 to June 2015 and explore its possible connection with the parsec-scale jet
evolution. We identify an escalating periodic oscillation (EPO) in the $\gamma$-ray light curve, characterized by a sequence of successive flux enhancements with progressively increasing temporal separations. Combining
Fermi-LAT data with contemporaneous 43 GHz Very Long Baseline Array (VLBA) observations, we find a temporal association between the EPO and the downstream propagation of a superluminal
knot. A helical trajectory model simultaneously fitting the $\gamma$-ray variability and the two-dimensional VLBA component positions reproduces the observed flux modulation and component motion. These results suggest that
geometric Doppler modulation caused by the changing viewing angle of a moving jet component provides a possible explanation for the observed EPO. This work provides a framework of connecting blazar variability to the
knot evolution of relativistic jets.
Read the full abstract ▾
All authors Zhihao Ouyang, Jingyu Wu, Hubing Xiao, Zhijian Luo, Zhu Chen, Shaohua Zhang, Liang Chen, Jianzhen Chen, Marina Manganaro, Luis C. Ho, Junhui Fan, Chenggang Shu
Data Structures and Algorithms
Strongly Refuting Semirandom Linear Systems in Subexponential Time
Pravesh K. Kothari, Andrew D. Lin and Peter Manohar
In this paper, we consider the problem of refuting $\mathbb{F}_2$-linear equations with random right-hand sides. Formally, we give a sub-exponential $2^{O(n/\log n)}$-time randomized
algorithm that takes as input an arbitrary $m \times n$
matrix $A$ and a uniformly random vector $b \in \mathbb{F}_2^m$, and outputs a witness showing that no assignment satisfies more than a $\frac{1}{2}+\epsilon$ fraction of the equations provided that $m \geq 2^{O(n/\log n)}$. The setting above is the semirandom refutation variant of the famous work [BKW03] that gives a $2^{O(n/\log n)}$-time search
algorithm for the learning parity with noise (LPN) problem with $m \geq 2^{O(n/\log n)}$ equations. Building on the search
algorithm of [Lyub05], we also give a $2^{O(n/\log \log n)}$-time refutation
algorithm that succeeds with only $m \geq n^{1 + \gamma}$ equations, for a small constant $\gamma$. Finally, we prove that our
algorithm is not captured by the sum-of-squares hierarchy by proving a degree-$\Omega(n)$ sum-of-squares lower bound, showing that ”[BKW03]-style”
algorithms achieve better runtime than can be done under sum-of-squares. We thus obtain a natural example of a noise-tolerant signal recovery problem that exhibits a nontrivial gap between the performance of efficient
algorithms and that of those based on the sum-of-squares hierarchy.
Read the full abstract ▾
Solar and Stellar Astrophysics
DKIST unveils MHD-predicted sub-50 km photospheric vortices
S. Vargas Dominguez, O. A. Calvo Rebellon, J. S. Castellanos Duran et al. (+4 more)
Solar vortices are rotating
plasma structures that play a key role in the transport of energy and
magnetic helicity from the photosphere to the chromosphere and corona. While numerical simulations predict the existence of small-scale photospheric
vortices with diameters of $50$–$100$ km, no direct observational evidence has been reported to date, primarily because of the limited spatial resolution of previous instrumentation. Here, we analyze high-resolution observations obtained with the 4-meter Daniel K. Inouye
Solar Telescope (DKIST) at 416 nm, with a spatial resolution of 12.3 km. Using
Fourier local correlation tracking and the swirling strength criterion ($\lambda_{\rm ci}$), we identify 201
vortex detections across a sequence of 143 reconstructed frames, spanning approximately 3.2 minutes, within a field of view of $4.92 \times 4.13$ Mm. The observations cover a photospheric plage region containing pore-like
magnetic concentrations. The detected
vortices have a median equivalent diameter of 38.6 km and a mean diameter of $39.5 \pm 3.3$ km. Virtually all detections are smaller than 50 km and therefore fall within the subgranular regime predicted by magnetohydrodynamic simulations but never previously observed. The median swirling period, $\tau_{\rm ci}=37.3$ s, satisfies the $\tau_{\rm ci}<100$ s criterion established by MURaM simulations. No statistically significant preference is found for prograde rotation (53.7%) over retrograde rotation (46.3%), consistent with the negligible influence of the Coriolis force at subgranular scales. These results demonstrate the unique capability of DKIST to resolve the population of photospheric convective
vortices predicted by state-of-the-art magnetoconvection simulations, opening a new observational window into small-scale
solar convective dynamics.
Read the full abstract ▾
All authors S. Vargas Dominguez, O. A. Calvo Rebellon, J. S. Castellanos Duran, M. van Noort, F. Wöger, D. A. Rodriguez Torres, C. G. Bernal
Artificial Intelligence
NNV3: Expanding Neural Network Verification to New Architectures and Domains
Anne M. Tumlin, Samuel Sasaki, Ben Wooding et al. (+8 more)
We present NNV3, the latest version of the Neural Network Verification (NNV) tool, a MATLAB framework for formal verification of deep learning models and learning-enabled cyber-physical systems. Building on the set-based reachability foundation of NNV 1.0 (FFNNs, CNNs, NNCS) and NNV 2.0 (RNNs, SSNNs, neural ODEs), NNV3 introduces new members of the
Star-set family: ModelStar for verifying networks under weight perturbation, VolumeStar for
video and 3D volumetric inputs, and GraphStar for graph neural networks. A conformal-inference-based probabilistic reachability mode complements sound analysis for problems where deterministic verification is intractable, while FairNNV certifies counterfactual and individual
fairness properties over continuous input regions. NNV3 introduces new
benchmarks for malware detection, graph-based power-system models, medical imaging, variable-length time series data, and action recognition. NNV3 also incorporates tutorials and developer guides through a unified documentation site. This paper details these major updates, demonstrating NNV's maturation into a comprehensive, robust, and accessible verification tool for a diverse range of AI systems.
Read the full abstract ▾
All authors Anne M. Tumlin, Samuel Sasaki, Ben Wooding, Diego Manzanas Lopez, Muhammad Usama Zubair, Navid Hashemi, Hongchao Zhang, Waseem Abbas, Ipek Oguz, Meiyi Ma, Taylor T. Johnson