Discrete Mathematics · Cross-listed from math.CO
Uniformly Weighted Graphical Designs
Zawad Chowdhury and Rekha R. Thomas
A graphical design is a subset of vertices of a graph, along with a weight for each chosen vertex, that can perfectly average chosen subspaces of functions on the graph. A design is uniformly weighted if all the weights are equal, and several well-known combinatorial objects such as orthogonal arrays, combinatorial block designs and t-wise permutations are uniformly weighted graphical designs. While one might expect to see uniformly weighted designs in structured graphs, they do not always exist. In this paper we characterize the existence of uniformly weighted graphical designs, and use our result to provide several families of graphs that have, and do not have, such designs. Our results offer a polyhedral view of the structures that control the existence and cardinalities of these designs. In particular, we characterize all uniformly weighted designs of threshold graphs, and provide a
geometric proof of the duality of linear codes and linear orthogonal arrays. We also provide a novel construction for graphs whose
Laplacian characteristic
polynomials are almost irreducible, to produce families without uniformly weighted designs.
Read the full abstract ▾
Systems and Control
Competitive One-Step-Ahead Control of Friedkin–Johnsen Networks: Potential Games, Stability, and the Price of Competition
Gabriel Gentil and Amit Bhaya
This paper studies competitive one-step-ahead control of Friedkin-Johnsen networks with overlapping player influence. The one-step interaction is an exact potential game with a unique Nash
equilibrium obtained from a symmetric positive-definite system. Sequential best-response sweeps converge for every frozen network state, parallel sweeps obey an exact Jacobi condition (and always converge with two players), and one-sweep implementations require an augmented state-action stability test. For marginal networks, a signed left-right damping condition is sufficient for exact-
equilibrium stability and becomes a sharp first-order instability test when its sign is reversed. A resolvent identity clarifies the feedback geometry, while a control-aware centrality identifies the goal conflicts that matter most. We characterize attainable
equilibria under unconstrained, convex, and sparse goal restrictions and give a closed form for the same-state welfare loss caused by competition. Numerical examples verify the stability thresholds, geometry, and welfare predictions.
Read the full abstract ▾
Software Engineering
Predicting LLM Performance from Prompt Linguistic Features: An Empirical Study in Requirements Engineering
Quim Motger, Alessio Miaschi, Xavier Franch et al. (+2 more)
Background. LLM outputs are highly sensitive to
prompt formulation: small wording changes can substantially affect output quality. This matters in software engineering, where
prompts guide requirements analysis, code generation, and artefact synthesis. Poor formulations yield unreliable artefacts, yet practitioners lack principled ways to assess a
prompt before inference, making selection depend on costly LLM calls and trial-and-error refinement. Aims. We investigate whether measurable linguistic properties of
prompts can predict LLM performance before inference, enabling low-cost
prompt selection and refinement, validated on binary requirements classification targeting F1, F2, precision, and recall. Method. We generate 9,000 linguistically controlled
prompt variants from 100 initial
prompts by varying 30 linguistic metrics, evaluated with five open-source LLMs on 625 annotated requirements. Regression predictors are trained via stratified 10-fold cross-validation with permutation-based significance testing; feature importance analysis identifies cross-LLM and model-specific predictors. Results. Linguistic features significantly predict
prompt performance across all targets (R2 in [0.38,0.42], q<0.05). Syntactic and morphosyntactic features drive most predictive signal; cross-LLM predictors include compound dependency distribution, conjunction density, and word/sentence length, reflecting sensitivity to domain vocabulary and complex structures. Conclusions. Results suggest practical implications for
prompt engineering, including overlap between linguistic patterns that reduce LLM performance and those that increase human comprehension difficulty, and the irrelevance of lexical variety as a quality dimension. More broadly, linguistic profiling combined with standard regression provides an effective, interpretable, low-cost prior before costly optimisation pipelines.
Read the full abstract ▾
All authors Quim Motger, Alessio Miaschi, Xavier Franch, Mohammad Amin Zadenoori, Alessio Ferrari
Graphics
Visual Cue Interactions in AR-Guided Needle Insertion: A Prostate Biopsy-Inspired Phantom Study
Xinrui Zou, Mingxu Liu, Thomas T. Jones et al. (+5 more)
Despite the apparent simplicity of the motor action involved during percutaneous needle procedures, manipulating the tool's direction becomes challenging when clinicians cannot directly visualize internal anatomy and must rely on ultrasound images, which increase cognitive demand. Augmented reality (AR) offers the promise to assist with these tasks by providing pertinent visual information in the clinician's field of view. However, simply providing visual information that ignores meaningful visual cues can complicate depth perception and spatial understanding. In this work, we introduce and evaluate three visualization techniques for needle
alignment developed during design sessions with medical experts: a localized focus-and-context window, a color-based proximity encoding, and an explicit trajectory overlay. These techniques were evaluated in a user study (n=26) including clinical experts (n=7) using a prostate-biopsy-inspired phantom. Results from this study suggest that cue effects depended on the surrounding cue configuration and user expertise. For novices, explicit trajectory overlay improved targeting accuracy and reduced retreat behavior, but its effect on completion time varied across cue configurations, with slower performance when the overlay was presented alone. For experts, the focus-and-context window reduced completion time and retreat events, while color-based proximity overlay improved completion time. Subjectively, color cues were often perceived as helpful even when their effects on accuracy were not consistent. These results suggest that AR guidance strategies for percutaneous interventions should consider user expertise and visual context.
Read the full abstract ▾
All authors Xinrui Zou, Mingxu Liu, Thomas T. Jones, Braden Millan, Sandeep Gurram, Peter A. Pinto, Raisa Z. Freidlin, Alejandro Martin-Gomez
Data Structures and Algorithms
Diva++: Dynamic Range Filtering over Hard Workloads
Navid Eslami, Ioana O. Bercea and Niv Dayan
Range filters are compact probabilistic data structures that answer approximate range emptiness queries. They are used in many domains, e.g., in key-value stores, to quickly rule out the existence of keys in a given query range and avoid searching for them in storage. However, all existing range filters exhibit at least one of three shortcomings: (1) they do not provide any false positive rate or performance guarantees, (2) they do not support variable-length keys and query ranges, and (3) they do not allow dynamic updates. We introduce Diva, the first range filter to address all the above challenges simultaneously. Diva learns the
dataset's distribution by sampling keys and storing them in a cache-efficient trie. It compresses keys in-between samples by removing their longest common prefix and truncating their suffixes while leaving enough bits in the middle (i.e., an infix) to differentiate the keys in sorted order. It stores infixes in constant-time dynamic data blocks, which it splits to handle insertions and expansions. It processes a range query by traversing the trie and checking for the inclusion of infixes in the target query range. We mathematically prove that Diva provides the best possible trade-off between
memory and false positive rate on many common real-world data distributions. We extend these benefits to a wider range of real-world workloads by introducing Diva++, an enhanced Diva variant. Diva++ saves
memory by removing redundancies among infixes using order-preserving
entropy encoding. It then removes any remaining identical infixes and uses the freed space to store more bits of the original keys within compact binary tries. We compare Diva and Diva++ to all prior range filters, and show that they achieve a false positive rate on par with the state of the art on real-world
datasets while supporting dynamicity and variable-length queries and keys.
Read the full abstract ▾
Social and Information Networks · Cross-listed from physics.soc-ph
Criticality and universality in network dismantling
Lorenzo Cirigliano, Claudio Castellano, Minsuk Kim et al. (+2 more)
Identifying the smallest set of elements whose removal dismantle a complex network, known as the network dismantling problem, is a fundamental task with many practical applications. Whereas network dismantling has been extensively studied over the past decade, most work has focused on developing efficient
algorithms for large but finite networks. By contrast, the physics of the network dismantling process, namely how the network structural connectivity is affected by the removal of nodes or edges, remains largely unexplored in the thermodynamic limit. Here, we shed light on this understudied aspect of network dismantling by introducing an adaptive biased percolation process able to optimally dismantle a network. Through a systematic analysis of synthetic network models, we find that the proposed percolation process displays a universal phase transition, characterized by the abrupt and simultaneous disappearance of both the
giant connected component and the largest 2-core, across networks with markedly different degree distributions. Simulations on real networks further support this universality, indicating that the physics of network dismantling is insensitive to a broad range of topological properties. Together, these results suggest that a
topology-agnostic theory could be developed to explain the critical behavior of network dismantling.
Read the full abstract ▾
All authors Lorenzo Cirigliano, Claudio Castellano, Minsuk Kim, Filippo Radicchi, Hanlin Sun
Neural and Evolutionary Computing
Tensor-Accelerated Eager Multi-Resolution Grids for Evolving Large-Scale Substrates
Romain Claret, Michael O'Neill, Paul Cotofrei et al. (+1 more)
In neuroevolution, indirect encoding generates neural network connectivity from a compact genome rather than specifying each connection. ES-HyperNEAT automatically discovers where to place hidden nodes by examining CPPN output patterns: it recursively subdivides space using a quadtree, expanding regions where CPPN outputs show high variance. This adaptive approach discovers network
topology without manual substrate specification, extending the fixed-grid HyperNEAT framework built on NEAT. However, the quadtree resists tensorization. Each depth level depends on the parent's variance, forcing sequential evaluation. Different CPPNs produce different subdivision patterns, preventing batching. And variable leaf counts are incompatible with JAX's static shape requirement for JIT compilation. Our prior work confirmed these limits at depths exceeding 5, and a JAX reimplementation of the quadtree yielded only marginal speedup despite batched optimizations, motivating the eager reformulation presented here. We present EMR-HyperNEAT, which evaluates all positions at all resolutions up front, then filters using the same variance criterion: ES-HyperNEAT's subdivide_if(var > $\theta$) becomes eval_all(); filter(var > $\theta$). This performs more CPPN queries than necessary, but all queries become independent and parallelizable across both cores and population members, reducing complexity from \BigO($4^D$) to \BigO($4^D/P$) across $P$ parallel cores. Recurrent substrate configurations become feasible through a connection type taxonomy. The experiments section validates 12-34$\times$ on-device GPU speedup on XOR at depths 5-7, and empirically higher solve rates across
benchmarks.
Read the full abstract ▾
All authors Romain Claret, Michael O'Neill, Paul Cotofrei, Kilian Stoffel
Computer Vision
ShiftSplit-AD: Separating Domain Shift from Defects in Foundation-Feature Visual Anomaly Detection
Muhamathu Ameer Ali Aacaas Muhamath
Visual anomaly detectors based on frozen foundation-model features commonly score distances from test patches to a
memory of normal features. Benign acquisition changes can also enlarge these distances, confounding domain variation with defects. We investigate whether structured decomposition of nearest-normal DINOv2 residuals can suppress shift-induced evidence while retaining unseen defects. ShiftSplit-AD decomposes the patch residual
matrix into low-rank and row-sparse components and scores the sparse component, with an optional low-rank/sparse fusion. The experiments expose a central trade-off rather than a universal separation: genuine defects can contain correlated, low-dimensional structure, so filtering broad residual activity may also remove defect information. On AeBAD-S, using settings fixed after Bottle development, sparse-only scoring improves image AUROC from 0.6780 to 0.7294 and AUPRC from 0.8052 to 0.8465. Paired bootstrap 95% intervals for the improvements are [0.0238, 0.0808] and [0.0170, 0.0650], respectively. However, sparse-only scoring reduces mean clean AUROC from 0.9890 to 0.9133 on four held-out MVTec categories and degrades Bottle localization. These findings show that residual decomposition can help when domain shift strongly contaminates anomaly evidence, but preserving defect structure remains the limiting problem.
Read the full abstract ▾
Robotics
Davood Soleymanzadeh, Kaidi Zhang, Zhiyuan Zhang et al. (+4 more)
Vision-
language-action models (VLAs) have shown strong promise for general-purpose robotic manipulation by mapping
language instructions and
vision observations directly to actions. However, most VLAs primarily condition action prediction on current observations and lack an explicit mechanism for
reasoning over future task dynamics, which is particularly important for fine-grained, contact-rich manipulation. We present PHR-VLA, a framework that enables planning-horizon
reasoning in VLAs through privileged latent representations of future dynamics. PHR-VLA introduces a lightweight auxiliary future head that, during training, aligns the VLA's internal representations with latent dynamics extracted from future observations. Evaluation results demonstrate that local, contact-centric, patch-level latent dynamics supervision from the wrist camera improves success rate on LIBERO from 84.1% to 88.4% and on real-world disassembly tasks from 63.3% to 82.5%. Patch-level supervision from a third-person camera also improves performance on Meta-World from 56.70% to 57.8%. These results demonstrate that privileged latent dynamics
alignment provides an effective training signal for improving anticipatory
reasoning in VLA policies. Project website: \href{https://davoodsz.github.io/PHR-VLA.github.io/}{https://davoodsz.github.io/PHR-VLA.github.io/}
Read the full abstract ▾
All authors Davood Soleymanzadeh, Kaidi Zhang, Zhiyuan Zhang, Bihao Zhang, Xiao Liang, Yu She, Minghui Zheng
Data Structures and Algorithms
Approximating the Rural Postman Problem below 3/2
Hong Li, Jianping Li, Wei Li et al. (+2 more)
We give an approximation
algorithm for the rural postman problem with approximation ratio strictly smaller than $3/2$. We obtain this result by adapting to the rural postman problem the technique of sampling from maximum
entropy distributions for the metric traveling salesman problem of Karlin, Klein, and Oveis Gharan. We also observe that, for every fixed $\varepsilon>0$, any $\alpha$-approximation
algorithm for the metric traveling salesman problem yields an $(\alpha+\varepsilon)$-approximation
algorithm for the rural postman problem; this implication is already implicit in the treatment of edges that must be traversed in the work of Lampis on the inapproximability of the traveling salesman problem.
Read the full abstract ▾
All authors Hong Li, Jianping Li, Wei Li, Runtao Xie, Xiaoxiao Yang
Game Theory
Approval-Based Apportionment: Like Portioning, Approximately like Committee Voting
Paul Gölz and Hannane Yaghoubizade
We study approval-based apportionment, a variant of committee elections in which candidates ("parties") can be selected several times. We show that the proportionality axioms EJR, EJR+, and FJR coincide, and so do PJR, PJR+, and FPJR; that Lindahl priceability, an axiom implying core stability, is equivalent to a notion of approximate optimality with respect to the proportional approval voting (PAV) score; and that locally PAV-optimal committees are priceable. Approval-based apportionment (where a candidate receives an integer number of seats) lies between committee elections (zero or one seat) and portioning (a fractional number of seats). We formally connect portioning and apportionment by giving a construction that lifts axioms from apportionment to portioning and preserves implications between them. Several of our new implications between apportionment axioms are natural from a portioning perspective, leading us to believe that apportionment sits closer to portioning than to committee elections. None of them holds in committee elections, but several extend approximately, which makes apportionment a fruitful setting for conjecturing approximate relationships in approval-based committee elections.
Read the full abstract ▾
Cryptography and Security
FlyBlind: Cross-Slice Timeliness Attacks on UAV Situational Awareness over 5G
Wagner Comin Sonaglio, Ágney Lopes Roth Ferraz, Sidnei Barbieri et al. (+3 more)
Beyond Visual Line of Sight (BVLOS) Uncrewed Aerial Systems (UAS) operating over 5G Standalone (SA) networks use a shared User Plane for both command-and-control (C2) data and
video feedback. Operators assess link quality through latency and availability, relying on soft isolation between network slices. However, the risk that an authorized co-tenant could make the Ground Control Station (GCS) state outdated without disrupting the connection remains underexplored. This work introduces FlyBlind, a timeliness attack in which an authorized co-tenant on a neighboring slice maintains legitimate uplink demand, causing state aging at the GCS without a rogue gNB or direct interference with C2 traffic. Our key insight is that, under soft isolation, sharing idle resources turns authorized competition for grants into state aging that conventional link monitors fail to detect. We formalize this effect, termed Silent State Staleness, as a falsifiable false-healthy predicate. On a dedicated testbed, telemetry age at the GCS saturates at approximately 12 seconds, with discrepancies of tens of meters between the GCS position estimate and ground truth, while the one-way delay (OWD) p99 remains in the tens of milliseconds, availability exceeds 99.9%, and the vehicle keeps operating in GUIDED mode without triggering failsafe mechanisms. These findings indicate that, in deployments with asymmetric uplink enforcement, verifying state freshness at the destination is essential rather than relying solely on link health.
Read the full abstract ▾
All authors Wagner Comin Sonaglio, Ágney Lopes Roth Ferraz, Sidnei Barbieri, André Elias Melo, Guevara Noubir, Lourenço Alves Pereira Júnior
Data Structures and Algorithms
Fine-Grained Complexity of Approximating Vector Knapsack: A Faster Algorithm and Bicriteria Optimality in 2D
Karl Bringmann, Ariel Kulik and Karol Węgrzycki
We revisit the $d$-dimensional Vector Knapsack problem ($d$-Knapsack): Given a $d$-dimensional capacity vector and a set of items, each with a $d$-dimensional weight vector and a profit, the goal is to select a set of items that maximizes the total profit without exceeding the capacity in any dimension. For any $d\ge2$, the best known approximation scheme for $d$-Knapsack runs in time $O(n^{\lceil d/\varepsilon\rceil-d})$ [Caprara, Kellerer, Pferschy, Pisinger '00]. We improve this running time to $\widetilde O_{d,\varepsilon,\rho}(n^{\lceil\frac{d-1}{2\varepsilon}-\frac12+\rho\rceil}+n^d)$ for any $\varepsilon\in(0,1)$ and every parameter $\rho\in(0,1)$. We achieve this speedup by designing the first meet-in-the-middle
algorithm for $d$-Knapsack. This requires replacing the LP solver used in prior
algorithms by a highly efficient dynamic programming
algorithm to generate representative solutions, building on an LP-based structural argument. This is the first improvement in over 25 years, and the first result that improves the exponent by a constant factor. We complement this by a fine-grained lower bound based on $k$-SUM showing that 2-Knapsack requires time $n^{\lceil\frac1{2\varepsilon}-\frac12 \rceil-o(1)}$. This establishes the optimal exponent of 2-Knapsack as $\frac1{2\varepsilon}\pm O(1)$, which is precise up to an additive $O(1)$. To the best of our knowledge, this is the first result that determines the optimal exponent more precisely than up to a factor $O(1)$, for any problem that admits a PTAS but no EPTAS. For the special case of 2-Knapsack we further attain a $(1-\varepsilon-\delta)$-approximation in time $\widetilde O_{\delta,\varepsilon}(n^{\lceil\frac1{2\varepsilon}-\frac12\rceil})$. This nearly matches our lower bound, as for a slightly better approximation ratio a slightly better running time is impossible – so our
algorithm is bicriteria-optimal.
Read the full abstract ▾
Formal Languages and Automata · Cross-listed from math.GR
Cycle Counting and Character Expectations Using Alternating Structures
Noam Ta Shma
Recently, two related papers [arXiv:2412.13941, arXiv:2409.03626] found a connection between two subjects: the w-cycle
theorem, which is a
theorem about counting appearances of cycles reading out a word w in certain graphs, and character expectations on word measures. The w-cycle
theorem was proven independently by [arXiv:1410.2540] using stackings and by [arXiv:1410.2579] using bislim structures. In the current work, we generalize stackings and bislim structures to alternating stackings and alternating bislim structures. We show how this significantly strengthens the w-cycle
theorem for words admitting such alternating structures, and as a result, also strengthens the recent results of [arXiv:2412.13941] and [arXiv:2409.03626]. We show that generic words admit alternating bislim structures, and therefore, the strengthened results hold for generic words. Using our new machinery, we address
conjectures of Wilton, of Hanany-Puder and of Puder-Shomroni. We prove that all three
conjectures hold for generic words, but we also find counterexamples for the first two.
Read the full abstract ▾
Emerging Technologies · Cross-listed from quant-ph
Plateau-Constrained Selection: Exploiting Degeneracy for Lower-Depth Quantum Compilation
Owen Friedewald, Ali Shiri Sichani and Chi-Ren Shyu
Minimum-cost orders of commuting phase terms can produce substantially different routed
circuits. On the same 36-term instances, three orders with identical support cost 74 yield mean routed depths of 228.6, 233.8, and 256.7. We exploit this degeneracy under fixed placement and maintained-parity lowering: Stage 1 attains the support optimum, and Stage 2 selects minimum routed depth among 24 equal-cost orders. For distinct pair supports, we characterize orders attaining the support lower bound through
Hamiltonian paths of the support line graph and count optima exactly through 20 terms. On synthetic 16-
qubit assignment-Ising instances, selection reduces depth by 12.83% under a different SABRE routing seed, with lower depth in all 20 instances. The selected orders lie a median 1.57 pool standard deviations below the pool mean, consistent with ordinary best-of-24 selection; the useful feature is that candidate rankings persist across routing seeds. Depth reductions extend to 48 terms and a random-MaxCut generator, whereas evaluation with BasicSwap reverses the gain. A 40-instance IBM Heron study measures a 0.59% error reduction on the executed stabilizer-probe panel, but the primary confidence interval across instances includes zero. Plateau selection therefore improves routed depth in the tested SABRE pipeline while preserving the logical support optimum.
Read the full abstract ▾
Machine Learning · Cross-listed from cond-mat.mtrl-sci
Physics-informed learning for the inverse problem in resonant ultrasound spectroscopy
Alejandro Cubillos Muñoz, Manuela Rivas and Julian Rincon
Inferring elastic constants from resonant ultrasound spectra is a nonlinear and typically overdetermined inverse problem based on finite spectral data. We formulate the Rayleigh-Ritz inverse problem as a constrained inverse-isospectral problem on the set of physically admissible elasticity tensors. This induces effective low-dimensional variables for the inverse map on the admissible elasticity
manifold: length and elastic scales, aspect-ratio coordinates, scale-free spectral features, and stability-respecting elastic ratios. We use these variables to construct a physics-informed learning pipeline in which a regression model acts only on reduced spectral and
geometric features, while scale recovery and final elastic-constant reconstruction are imposed analytically. For the full cubic
benchmark, the reconstructed constants have MAE values of $20.37(35.15)$, $24.30(41.33)$, and $2.13(3.66)~\mathrm{GPa}$ for $C_{11}$, $C_{12}$, and $C_{44}$. In the fixed-geometry
benchmark, the corresponding cubic MAPE values are $4.14(3.87)\%$, $8.31(8.50)\%$, and $2.44(2.86)\%$, while the isotropic values are $4.0(3.6)\%$ and $0.4(0.3)\%$ for the bulk and shear
moduli. The inverse problem then becomes a constrained regression problem in variables adapted to the geometry, scaling,
crystal symmetry, and thermodynamic stability of Hookean elasticity.
Read the full abstract ▾
Computer Vision
Quanta Perception as Probabilistic Events
Varun Sundar, Pavan Thodima, Sacha Jungerman et al. (+1 more)
Autonomous systems rely on extracting information from light, yet remain brittle in extreme environments, from nighttime
navigation to high-speed robotics. Conventional sensors aggregate
photons over fixed exposures, imposing trade-offs between sensitivity, dynamic range, and temporal resolution that degrade perception when
photons are scarce or dynamics are rapid. Quanta sensors detect individual
photons, but their streams exceed real-time compute and latency budgets by orders of magnitude. Here we introduce $\textit{probabilistic events}$, a computational primitive for real-time quanta perception from individual
photon detections. By computing the posterior over the time since the last intensity change, we represent
photon streams as recursive belief states. Rather than fixed-threshold event-camera triggers, this recursive
Bayesian formulation yields three low-latency signals: motion-adaptive scene flux, high-fidelity activity maps, and
entropy-based perceptual uncertainty. This representation enables perception in extreme conditions, including pose estimation of a running person at $\sim$0.05 lux—without retraining
vision models. Our approach processes input streams exceeding 50{,}000 quanta frames per second on commodity GPU hardware—yielding kilohertz-scale outputs up to four orders of magnitude faster than state-of-the-art quanta reconstruction baselines, even for megapixel arrays. By replacing frame reconstruction with direct probabilistic inference over
photon streams, this work bridges
photon-counting quanta sensing with robotic
vision.
Read the full abstract ▾
All authors Varun Sundar, Pavan Thodima, Sacha Jungerman, Mohit Gupta
Artificial Intelligence
LongGuard: Mechanistic Analysis and Training-Free Mitigation of Long-Context Failure in Safety Guardrails
Ziyang Chen, Xing Wu and Songlin Hu
Safety guardrails serve as the last line of defense against harmful inputs and outputs of large
language models (LLMs), yet they are trained and evaluated almost exclusively on short text. We present LongGuard, a framework that evaluates, mechanistically analyzes, and mitigates long-context guardrail failure. We formulate the task as Safety Needle-in-a-Haystack (SafetyNIAH) over a 0.25k-32k length grid; across 15 mainstream guardrails, unsafe recall drops monotonically by more than 50% on average, and a paired Benign-Fill vs. Needle-Repeat design attributes the failure to proportional dilution of the unsafe needle rather than to absolute length. A three-layer
attention-logit-behavior analysis on six guardrails locates the mechanism:
attention mass on the unsafe needle is diluted, the unsafe-over-safe logit margin is compressed in lockstep, and the detection decision collapses accordingly, with this
attention->logit->behavior chain remaining consistent after partialling out length. We further isolate a sparse set of guard-specialized
retrieval heads that exhibit partial specificity relative to their base models. Building on the analysis, we propose two training-free mitigations - Chunked Detection (CD) and
Attention-Head Sharpening (AHS) - and a deployment protocol, Context-Aware Hyperparameter Routing (CAHR), that selects configurations by context length and audit side. Across five
benchmarks spanning synthetic data, long-context attacks, and
reasoning-model outputs, CAHR-CD and CAHR-AHS improve the six-guardrail average by 22% and 13%, respectively. Code and data are available online.
Read the full abstract ▾
Machine Learning · Cross-listed from physics.plasm-ph
Towards Large-Scale Heterogeneous Data Organization for Scientific Foundation Models: A Nuclear Fusion Case Study
Nathaniel Chen, Kouroche Bouchiat, Peter Steiner et al. (+3 more)
Training effective foundation models requires massive and organized
datasets, yet scientific domains such as
nuclear fusion present unique challenges due to largely heterogeneous and sparse data. Here we characterize the data used in developing such a model: with over 20 sensor types spanning 5 orders of magnitude in sampling rate, mixed tensor structures (point measurements, spectrograms, images), and nonstationary physics. We analyze our input complexity and discuss trade-offs between temporal context and frequency resolution. Our analysis provides a template for representing multi-modal fluctuation data at scale, with implications for both multi-modal control systems and
nuclear fusion.
Read the full abstract ▾
All authors Nathaniel Chen, Kouroche Bouchiat, Peter Steiner, Azarakhsh Jalalvand, SangKyeun Kim, Egemen Kolemen
Machine Learning
Self-Explainable Multi-Label Graph Neural Network for Correlated Evidence Attribution
Yingqi Feng, Yufei Tang, Min Shi et al. (+1 more)
Multi-label graph learning intends to capture the intrinsic complexity of real-world applications, where one sample is often related to multiple groups or consists of multiple objects. To date, a handful of multi-label graph learning methods exist, but none of them integrate training-time interpretation capability. While post-hoc graph explainers have been developed, they do not explicitly model label-dependent evidence sharing in multi-label graph learners, especially when label pairs are weakly or negatively associated. As a result, post-hoc approaches may miss how evidence should be shared or separated across different labels. This paper advances a new end-to-end self-explainable multi-label graph neural network (SEMGNN), which aims to simultaneously classify multi-labeled nodes and identify edges significantly contributing to each target node w.r.t. predicted labels. Different from post-hoc methods, SEMGNN jointly learns a predictor and a sparse edge-mask explainer within a unified framework and training objective. Label-label correlations are used to improve multi-label node classification and enhance individual label explanations, so that different labels of a node can be supported by distinct yet coherent structural and/or correlated evidence. Experiments and comparisons on synthetic and real-world multi-label networks, in social networking, entertainment, and life sciences, show that SEMGNN achieves competitive or improved predictive performance while providing more faithful and compact label-conditioned explanations.
Read the full abstract ▾
All authors Yingqi Feng, Yufei Tang, Min Shi, Xingquan Zhu
Information Theory · Cross-listed from quant-ph
Spectral Theory of Semisimple Bivariate Bicycle Codes
Eric Sabo, Mahir Bilen Can and David Marquis
Extending the classical theory of two-dimensional cyclic codes, we develop an algebraic approach to bivariate bicycle codes. Using Frobenius-orbit idempotents, formulas for logical dimensions are derived and lower bounds on minimum distances are established. A systematic theory of code
symmetries is formulated to construct a structured block-monomial subgroup of coordinate permutations. Several explicit examples show how to generate these codes from first principles without relying on numerical searches. An appendix extends the analysis to BCH-based product constructions.
Read the full abstract ▾
Computer Vision
VidParse: Online Parsing of Egocentric Procedures Like a Pro
Anubhav Gupta, Archit Kambhamettu, Vatsal Agarwal et al. (+2 more)
Translating continuous, noisy
egocentric video streams into discrete, temporally ordered action steps is fraught with visual challenges. Heavy ego-motion, transient occlusions, and the high intra-class variability of unscripted human-object interactions cause standard frame-level online temporal models to struggle, often resulting in severe over-
segmentation and structural collapse. To bridge the gap between unstable low-level perception and high-level procedural logic, we present VidParse, an online, training-free framework that treats activity understanding as a graph-constrained inference problem. Rather than relying on learned temporal filters, we dynamically identify semantic transitions using a temporal similarity
matrix over manipulation-anchored features, which are extracted from frozen foundation models to prioritize foreground hand-object interactions. A beam search decoder then leverages an induced procedural task graph to explicitly enforce valid action transitions and prune impossible trajectories. By anchoring robust visual segments to hard procedural constraints, our approach preserves long-range state transitions and achieves up to a 10x improvement in complex multi-step parsing accuracy over strong online baselines, all without requiring a single gradient update.
Read the full abstract ▾
All authors Anubhav Gupta, Archit Kambhamettu, Vatsal Agarwal, Pulkit Kumar, Abhinav Shrivastava
Robotics
Beyond Data Scaling: Representation-Centric Continued Pre-training for Vision-Language-Action Models
Senqiao Yang, Chengyao Wang, Yuxin Chen et al. (+13 more)
Scaling
robot data is crucial for building generalist
Vision-
Language-Action (VLA) models, yet
robot trajectories are harder to scale than web-scale image-text data because
embodied collection is costly and sparsely covers the physical world. This makes representation quality a central bottleneck: under a fixed
robot-data budget, continued pre-training must turn limited trajectories into transferable visual-action knowledge rather than merely fit actions. We propose VLAct, a VLA-oriented VLM backbone trained on broad, heterogeneous, multi-embodiment
robot data before task-specific fine-tuning. VLAct preserves the broad VLM prior and encourages shared action semantics across embodiments through VLM-prior preservation, multi-head continuous action co-supervision, and a partially unified cross-embodiment action layout, while allowing task-specific action heads during fine-tuning. Across simulation, real-world, and unseen-embodiment transfer, VLAct consistently improves downstream performance under fixed fine-tuning protocols. On LIBERO-Plus and RoboTwin 2.0, VLAct surpasses industrial VLA systems including ABot-M0 and LingBot-VLA, achieving success rates of 82.6% and 92.5%. On RoboDojo, VLAct ranks sixth among all policies by success rate and outperforms all explicitly designated world-action model (WAM) entries on both metrics. Most notably, on RoboCasa-GR1, an unseen
humanoid embodiment, VLAct using only 20% of downstream trajectories outperforms the full-data GR00T-N1.6 baseline. These results are obtained using fully open-source data and only a 16-GPU training setup, showing that representation-centric continued pre-training can deliver highly competitive performance under a modest compute budget and is an important independent axis of VLA progress beyond data scaling.
Read the full abstract ▾
All authors Senqiao Yang, Chengyao Wang, Yuxin Chen, Zixuan Wang, Longxiang Tang, Haokun Gui, Jinhui Ye, Changsheng Lu, Xiaoyang Wu, Mingkang Zhu, Pengguang Chen, Shu Liu, Zhuotao Tian, Hengshuang Zhao, Bei Yu, Jiaya Jia
Computer Vision
Code as Worlds: Agentic Discovery of Executable World Representations for Physical Reasoning
Hanyang Wang, Yimo Cai, Weiliang Chen et al. (+14 more)
Physical understanding and
reasoning depend on forming compact and generalizable representations of the world. While modern
vision-
language models can recognize and explain diverse physical events, they often lack explicit representations of the underlying mechanisms-such as object states, physical parameters, and governing dynamics-needed for reliably
reasoning how the world evolves and responds to interventions. In this work, we introduce Code-as-World, a paradigm that represents physical worlds through executable world representations. By expressing physical composition, dynamic
evolution, and visual appearance as executable code, Code-as-World provides a compact, quantitatively grounded, and controllable abstraction of the physical world. To construct such representations from
multimodal observations, such as natural-
language descriptions or real-world
videos, we develop an agentic discovery loop inspired by abductive
reasoning, where an
agent proposes, executes, renders, verifies, and iteratively refines executable world hypotheses. As a concrete application, we use verified executable worlds to provide scalable physical supervision for training
vision-
language models on quantitative physical
reasoning. Experiments show that Code-as-World-VL achieves state-of-the-art performance on QuantiPhy and surpasses leading proprietary models, highlighting the potential of executable world representations as a scalable foundation for physical intelligence.
Read the full abstract ▾
All authors Hanyang Wang, Yimo Cai, Weiliang Chen, Jiawei Chi, Haowen Sun, Qiyu Dai, Yi-Hsin Hung, Xingzhuo Guo, Jinshan Ren, Runmao Yao, Ziwei Liu, Mingsheng Long, Yueqi Duan, Jun Gao, Jiangran Lyu, Fangfu Liu, Jialong Wu
Artificial Intelligence
Nemotron 3.5 Content Safety Moderator: A Compact Multimodal, Multilingual, and Reasoning Enabled Content Safety Moderator
Varun Singh, Anuj Doshi, Makesh Narsimhan Sreedhar et al. (+2 more)
Safety moderation for deployed AI applications is moving beyond text-only
prompts: systems increasingly need to judge images, documents, screenshots, and generated responses under policies that vary across domains. Existing guardrails usually cover only part of this setting, making it difficult to combine broad coverage, custom policy control, and low compute cost. We present Nemotron 3.5 Content Safety Moderator, also referred to as Nemotron 3.5 CS in this paper for brevity, a compact 4B
vision-
language safety moderator that jointly classifies user
prompts, images, and assistant responses across 12
languages. Nemotron 3.5 CS returns safety labels for latency-sensitive moderation and can additionally produce concise
reasoning traces that apply supplied custom policies and identify violated categories when
reasoning is requested. We also release a
multimodal and
multilingual safety
dataset for guard training, spanning human-labeled real-image moderation, benign
vision-
language and document tasks, synthetic rare-risk and jailbreak cases, and custom-policy examples. Across evaluations spanning
multimodal safety, text moderation,
multilingual robustness, custom-policy following, benign false positives, and latency, Nemotron 3.5 CS demonstrates a practical coverage tradeoff: it adds image-conditioned and policy-conditioned moderation while remaining broadly competitive with specialized guard models. These results suggest that compact
vision-
language moderators can serve as deployable front-line safety components, with
reasoning used selectively for audit and policy review.
Read the full abstract ▾
All authors Varun Singh, Anuj Doshi, Makesh Narsimhan Sreedhar, Shaona Ghosh, Katherine Luna
Robotics
Remote Human-Robot Interaction In Greenhouses via Virtual Reality: How Plant Canopy Structure Affects Leaf Disease and Soil Moisture Inspection
Daniel Udekwe and Hasan Seyyedhasani
This study evaluates the effectiveness of remote human-
robot interaction using virtual reality for leaf inspection and soil moisture assessment in a greenhouse environment. The robotic system comprised an unmanned ground vehicle and a robotic manipulator equipped with cameras, governed by kinematic models for
navigation and manipulator control. Fourteen distinct plants were inspected across two experiments utilizing VR teleoperation, guided by a set of pre-specified research questions and hypotheses. In the leaf inspection experiments, cycle completion times varied from 3.3 to 8.0 s, and plant-based disease detection was achieved up to 88% accuracy; diseased-spot detection improved numerically in the second experiment, though this change was not statistically significant (p=0.378). For soil moisture assessment, the experiments achieved successful determination of watering needs in up to 64.3% of plants (9 of 14), with consistent success observed for plants 1, 2, 3, 8, 9, 10, and 13; however, this improvement was likewise not statistically significant (p=0.50). A post hoc analysis instead revealed that soil moisture assessment reliability was strongly and significantly predicted by plant canopy morphology (p<0.01): plants with broad, single-leaf canopies reached 100% success by the second experiment, versus only 16.7% for dense, compound canopies. A secondary analysis showed operators became measurably faster at attempting dense-canopy plants without a corresponding gain in success, indicating that camera occlusion, not operator skill or effort, is the dominant limiting factor. These findings show occlusion imposes a sensing limitation rather than a control or training deficiency, and that adapting camera viewpoint and sensing strategy to canopy density is needed to improve the system's accuracy and robustness.
Read the full abstract ▾
Information Theory · Cross-listed from stat.ML
Towards a mathematical theory of superposition
Michael I. Ivanitskiy, John Jasper, Emily J. King et al. (+1 more)
We develop a mathematical theory of superposition in neural networks using tools from frame theory and compressed sensing. In our model, a sparse binary vector \(x\) of active features is encoded through an overcomplete dictionary \(W\), and feature recovery is performed by applying \(\operatorname{ReLU}(W^\top W x+b)\) with an appropriate bias vector \(b\). We prove several recovery
theorems for this model. In the random-support setting, we establish high-probability support recovery for nearly tight, low-coherence dictionaries, with guarantees when the expected sparsity is up to order \(d/\log n\). In the worst-case support setting, we give a sharp and computable criterion for which sparsity levels permit support recovery. We apply this criterion to Gaussian random
matrices and equiangular tight frames. For real equiangular tight frames with \(n>d+1\), we determine the exact recovery threshold in terms of the coherence. The proof of this result for real equiangular tight frames relies on a novel characterization—which should be of independent interest to frame theorists—of the distribution of signs in the Gram
matrix.
Read the full abstract ▾
All authors Michael I. Ivanitskiy, John Jasper, Emily J. King, Dustin G. Mixon
Social and Information Networks
Why People Share Social Media Screenshots
Tarannum Zaki, Michael L. Nelson and Michele C. Weigle
Posting screenshots of social media posts is a common way of increasing platform functionality, enabling inter-platform interoperability, and gives users more control over how content is presented and interpreted. While these are legitimate purposes, screenshot sharing is also used in ways that can remove context, hide the original source, or mislead audiences. We examine the motivations behind sharing screenshots of social media posts on social media platforms like Twitter/X, Instagram, and Facebook. By focusing on user motivations, our study contributes to understanding how screenshot practices shape the sharing and interpretation of information in online spaces.
Read the full abstract ▾
Game Theory
Refundable Deposits: How to Restore Cooperation in Finitely Repeated Games
Giulio Salizzoni, Domenico Mergoni Cecchelli, Edward Plumb et al. (+2 more)
While infinitely repeated games admit a rich set of Nash
equilibria, finitely repeated games typically have a much smaller and often inefficient one. We show how to enlarge this set using deposits: in each period a player may place a refundable sum with a neutral intermediary, returned when the game ends and forfeited following a deviation. Paying these deposits is voluntary and incentive compatible at every stage, so no commitment by the players is assumed, the only commitment required being that of the intermediary to a refund rule fixed before play begins. The mechanism sustains payoff profiles more efficient than those of the standard
equilibria, without altering the underlying game and without transfers between players. We demonstrate it on the prisoner's dilemma, a congestion game, and a public goods game, all settings where cooperation cannot emerge in the standard finitely repeated version. We also apply it to a dynamic common-pool resource, suggesting that the construction extends beyond repeated stage-games.
Read the full abstract ▾
All authors Giulio Salizzoni, Domenico Mergoni Cecchelli, Edward Plumb, Maryam Kamgarpour, Galit Ashkenazi-Golan
Machine Learning · Cross-listed from cond-mat.mtrl-sci
Ab initio Modeling of MoS2/Oxide Device Interfaces with Machine Learned Electronic Structures
Manasa Kaniselvan, Mauro Dossena, Denghui Lu et al. (+4 more)
We introduce a new ab initio approach to simulate
semiconductor devices that integrates scalable machine-learned (ML) electronic structure models with an advanced
quantum transport (QT) solver. The developed framework enables 10,000X speedups over density functional theory to produce the
Hamiltonian matrix of devices made of >20,000 atoms, while offering high prediction accuracy. We use its unique features to investigate MoS2/
oxide samples and single-layer MoS2 field-effect transistors, where the surrounding
oxide layers, here, HfO2 or Al2O3, are explicitly included into the QT domain. In particular, we reveal that the presence of undercoordinated metal atoms (Hf or Al) close to the
semiconductor-
oxide interface significantly affects the magnitude of the electronic current and its propagation through MoS2.
Read the full abstract ▾
All authors Manasa Kaniselvan, Mauro Dossena, Denghui Lu, Alexander Maeder, Nicolas Vetsch, Alexandros Nikolaos Ziogas, Mathieu Luisier