arXiv subjects
Search indexed arXiv papers on artificial intelligence, large language models, computer vision and robotics. Read source abstracts and follow links to arXiv.
If the machines of some type A have enough resources to (i) solve problems that no machine of type B can solve, and (ii) simulate any unary two-way deterministic finite automaton that has access to a linearly-bounded counter, then typically no recursive function can upper bound the increase in the size of description when a machine of type A is replaced by an equivalent machine of type B.
Exact statements about how a system far from equilibrium responds to changes in its rate constants are scarce beyond linear response, yet they are central to the control of biochemical networks. A notable exception is mutual linearity: in a jump process, controlling the rates of a single transition makes any two stationary probabilities obey an exact affine relation, arbitrarily far from equilibrium. With nonlinear mass-action kinetics, and conservation laws in place of a normalization, chemical reaction networks seem to defy mutual linearity since species concentrations obey no such relation. We show that it was there all along, in the complexes' activities (mass-action monomials) rather than in the concentrations of species. For zero-deficiency networks with a single linkage class, controlling one reaction makes any three stationary activities obey a linear relation whose coefficients are independent of the controlled rate constants, the conserved quantities (hence the initial state), and even of the stoichiometry of complexes. This predicts hidden activities and bounds the activity ratios the control can reach, while a violation signals nonzero deficiency or multiple linkage classes.
Vision-language-action (VLA) policies assume the embodiment on which they were trained and can fail when a joint fault changes how commanded actions are physically executed. Existing fault-recovery methods often require task-specific retraining, fault labels, explicit diagnosis, or privileged embodiment information. We introduce CAPABLE, a unified capability-aware adaptation framework for frozen VLAs that integrates self-supervised capability inference with residual reinforcement learning. CAPABLE infers capability, how much of the commanded motion each joint actually realizes and how that motion contributes to end-effector behavior, online from command-response history and kinematics using a temporal encoder shared across joints, Jacobian grounding, cross-joint attention, and self-supervised physical prediction. The resulting representation conditions a residual policy that adds bounded corrections to the VLA arm action without fault labels or faulty-joint identifiers. Across 28 LIBERO tasks, CAPABLE raises success on an actuator excluded from fault training from 24.8% to 59.3%, outperforming a parameter-matched global-history baseline by 17.4 points while preserving healthy performance. Leave-one-actuator-out experiments across six joints show that this transfer is not specific to one actuator, and additional evaluations characterize transfer to unseen fault families and demonstrate recovery on a physical Franka Panda. https://capable-vla.github.io/
In this work, we study the fast Fourier transform (FFT) of discrete data with crystallographic symmetry, for which standard FFT algorithms incur substantial redundant computation. We establish a rigorous mathematical theory of the symmetric fast Fourier transform (SFFT) for $n$-dimensional crystallographic groups and propose the SFFT and symmetric inverse fast Fourier transform (SIFFT) algorithms. We derive symmetry-reduction formulas for both linear and translational symmetries and unify them into a recursive framework. We analyze the arithmetic complexity to obtain theoretical speedup factors relative to the standard FFT, and we confirm the accuracy and efficiency of the proposed methods by numerical experiments on representative three-dimensional space groups.
Integrating over-the-air computation (AirComp) with non-orthogonal multiple access (NOMA) enables concurrent data aggregation and information transmission, but tightly couples computation accuracy with communication quality of service (QoS). This paper investigates a pinching-antenna-assisted uplink, where flexible pinching-antenna (PA) placement provides additional spatial degrees of freedom to manage this coupling. We minimize the AirComp mean-squared error (MSE) subject to NOMA QoS, transmit-power, and PA deployment constraints by jointly designing the AirComp transceiver, PA positions, and successive interference cancellation (SIC) order. Exploiting the recursive structure of uplink SIC, we establish an exact NOMA feasibility condition and derive an optimal SIC-ordering rule. These structural results eliminate the NOMA power variables and the combinatorial search over decoding orders, reducing the original problem to a continuous optimization problem. We then develop an iterative algorithm with optimal AirComp transmit-coefficient and receive-scaling updates and a feasibility-preserving majorization scheme for PA placement. The proposed algorithm preserves NOMA feasibility while generating a monotonically nonincreasing computation-MSE sequence. Numerical results demonstrate the computation-accuracy gains of jointly optimizing the AirComp transceiver and PA positions under NOMA QoS requirements.
Let $S_k$ be the space of cusp forms of weight $k$ for ${\rm SL}_2(\mathbb Z)$, with $\dim S_k\geq1$, and let $χ_p=\left(\frac{\cdot}{p}\right)$ for odd primes $p$. For fixed $s$ with ${\rm Re}(s)>(k+1)/2$, we prove that the proportion of primes $p\leq X$ satisfying $L(h,χ_p,s)=0$ tends to zero as $X\to\infty$, uniformly over nonzero $h\in S_k$. We also prove an analogous statement for nontrivial linear combinations of twisted symmetric square $L$-values of the normalized Hecke eigenforms when ${\rm Re}(s)>k$. The main ingredient of the proof is a non-concentration theorem for random Euler products under a pairwise local separation condition. As applications, we obtain spanning and density-one basis results in $S_k$ for the kernels associated with twisted standard and symmetric square $L$-functions, with explicit basis criteria in specified right half-planes when $\dim S_k=2$ or $3$. We show that almost every tuple of $\dim S_k$ primes gives a basis of $S_k^{\ast}$ from twisted periods at every noncentral critical index. We also prove a density-one result for rational bases of $S_k$ obtained from traces of Rankin-Cohen brackets of Eisenstein series at prime levels.
Learned simulators for deformable linear objects (DLOs) such as cables have to predict the motion of cables they were not trained on and stay stable over long rollouts. Most of their error occurs where the cable touches itself or the floor. Attention over all pairs of cable segments can represent contact between parts of the cable that are far apart along its length, but attention has no notion of geometry. A cable has two pairwise distances, which agree only while it is straight: the arc-length distance along the cable, which governs elastic forces, and the Euclidean distance in space, which governs contact. We add a physical attention bias, an additive term on the attention logits with a learned rate, and ask which distance it should use. We compare no bias, each distance alone, and both distances on disjoint sets of heads, keeping the rest of the model and the training protocol fixed. A physical bias improves prediction on unseen cables. The gain is largest when attention is the only mechanism that connects distant segments: there, the arc-length bias reduces prediction error by 15% and more than halves the drift in segment length. The Euclidean bias alone stays close to unbiased attention, while assigning both distances across heads is best or near-best on every metric we report. Code and per-run records: https://github.com/avihaig/dlogps.
We detail scalable methods for approximating the quadratic entropy $p^T d p$ for arbitrary distributions $p$ and common distances $d$ of negative type. We focus on the Euclidean and spherical geodesic cases, which both use random feature embeddings and projections to dramatically improve computational complexity within a simple framework. Amortization of a single large matrix multiplication and control variates further enable computation at large scale with low memory and runtime in situations where $d$ is held constant while $p$ varies. We demonstrate this with a comparison against direct pair sampling and bibliometric/scientometric examples on Open Graph Benchmark datasets, revealing papers, fields, and institutions with both particularly narrow and broad interdisciplinary reach from their citations and text features alone.
Organic mixed ionic-electronic conductors (OMIECs) couple ionic and electronic signals, enabling energy storage, neuromorphic computing, and bioelectronics. However, how charge relaxes in water once the bias is removed, which determines how long stored states persist, remains unclear. Using operando charge photometry, we tracked charge in OMIEC films during biasing and open-circuit relaxation. Under bias, charges moved as redox fronts. After bias removal, charge instead relaxed uniformly across the channel. Modeling and cryo-electron microscopy indicate that swelling creates connected water-rich pathways through the bulk. These pathways speed ion motion but also let stored charge leak away. Tuning the terminal side chain of a polythiophene controls water uptake and ion trapping. Bulky hydrophobic side chains enable stable programmable states, whereas hydrophilic side chains support fast electrophysiology sensing.
Jev is a "System One" model that returns a choice among given options instead of generating text. We study how such a specialized decision model compares with general-purpose large language models (LLMs). We evaluate Jev on 13 multiple-choice benchmarks covering knowledge, reasoning, and multilingual understanding, and compare it with 19 LLMs in three tiers: frontier, representative, and small. Jev is competitive with frontier LLMs on knowledge and commonsense benchmarks and obtains the best score on MMLU-Redux and ARC-Challenge. Outside mathematics, it also outperforms most representative LLMs and all small LLMs. However, it falls behind on mathematical word problems: on MathQA, it is 17.7 points below the frontier median and scores lower than all 19 LLMs. These results indicate that a specialized decision model can match general-purpose LLMs on decisions that rely mainly on knowledge, but not on decisions that require multi-step calculation.
In this paper, we study the tensor triangular geometry of band algebras $RB$, where $B$ is a finite left regular band and $R$ is a commutative Noetherian ring. We show that the Balmer spectrum of $\Perf(RB)$ is homeomorphic to $\Spec R\times L_B$, where $L_B$ is the support lattice of $B$. We also show that there is a bijection between the localizing ideals of $D(RB)$ and the stable subsets of $\Spec R\times L_B$. As a consequence, the telescope conjecture holds for $D(RB)$. Finally, we construct a band $B$ such that the generalized Nerves of Steel Conjecture fails for the non-rigid tensor triangulated category $\Perf(RB)$.
The profound mathematical and conceptual analogy between pressure in fluid mechanics and gravity in general relativity motivates a novel geometric reinterpretation of flows. This article proposes a theoretical framework for geometrizing pressure by establishing a Newton-Cartan geometry and applying a conformal transformation to the spatial metric. The pressure field is reinterpreted as a geometric potential that induces an effective spatial curvature, encoded in the Ricci tensor Rij and the associated Einstein-like tensor Gij. Within this curved effective space, fluid particles under the exclusive influence of pressure follow geodesics, and the classical Euler equation is recovered in the weak-field limit. The pressure Poisson equation emerges naturally as a consequence of the geodesic equation and the Bianchi identity, revealing pressure as an instantaneous geometric constraint rather than an external force. This framework provides a unified geometric interpretation of widely used vortex identification criteria, demonstrating that the prevailing Eulerian and Lagrangian schemes correspond to different geometric invariants or projections of the effective curvature. In the degenerate two-dimensional (2D) limit, the effective spatial tensor vanishes identically, and the global integral of the total curvature vanishes as well. This topological constraint offers a novel interpretation of the dual-cascade scenario in 2D turbulence: the inverse energy cascade arises from geometric condensation of positive-curvature structures toward larger scales, while the forward enstrophy cascade results from geometric dispersion of negative-curvature structures toward smaller scales.
By analyzing $(2712.4\pm14.3)\times10^6$ $ψ(3686)$ events collected with the BESIII detector operating at the BEPCII collider, we report the first observation of $χ_{cJ}\to ΛK^0_S\bar Ξ^0 + c.c.$ ($J=0,1,2$) with significances greater than $10σ$, including systematic uncertainties, and improved measurements of $χ_{cJ}\to ΛK^-\bar Ξ^+ + c.c.$ The branching fractions of these decays are determined to be $\mathcal{B}(χ_{c0}\to ΛK^0_S\bar Ξ^0+c.c.)=(1.13\pm0.13\pm0.13)\times10^{-4}$, $\mathcal{B}(χ_{c1}\to ΛK^0_S\bar Ξ^0+c.c.)=(0.74\pm0.08\pm0.07)\times10^{-4}$, $\mathcal{B}(χ_{c2}\to ΛK^0_S\bar Ξ^0+c.c.)=(1.16\pm0.11\pm0.14)\times10^{-4}$, $\mathcal{B}(χ_{c0}\to ΛK^-\bar Ξ^++c.c.)=(1.90\pm0.08\pm0.16)\times10^{-4}$, $\mathcal{B}(χ_{c1}\to ΛK^-\bar Ξ^++c.c.)=(1.40\pm0.05\pm0.11)\times10^{-4}$, and $\mathcal{B}(χ_{c2}\to ΛK^-\bar Ξ^++c.c.)=(1.95\pm0.07\pm0.15)\times10^{-4}$. The first uncertainties are statistical, and the second are systematic.
We investigate the effects of quantum geometry on the motion of magnon wave packets in one-dimensional quantum spin chains. In contrast to fermionic systems, where transport phenomena driven by the quantum metric have been extensively studied, we show that the dynamics in bosonic magnon systems are strongly affected by the symplectic structure and the pseudo-unitarity of the transformation of the magnon Bogoliubov-de Gennes Hamiltonian. A chiral soliton lattice provides a setting in which the symplectic quantum metric strongly influences the longitudinal magnon dynamics. We derive the semiclassical response of a magnon wave packet in such a system to a weak, slowly varying perturbation and demonstrate that the wave packet velocity acquires a geometric contribution proportional to the momentum derivative of the symplectic quantum metric, which dominates over the conventional group velocity for a nearly flat band. Solving the resulting equations of motion, we obtain a bounded trajectory in real time with a phase-space area determined by the symplectic quantum metric.
Bayesian optimisation is the natural tool for shape design when objectives are expensive and non-differentiable, but it needs a compact yet expressive parameterisation of the search space. Hand-crafting one is a complex endeavour requiring domain expertise, and often yields implicit infeasible regions, artificial bounds, and coupled, unordered coordinates. We instead learn the parameterisation from a collection of existing designs, applying principal component analysis to the deformations between shapes. The result is a linear, interpretable search space in which the number of components explicitly trades expressivity against dimensionality. Across aerofoils, wings, and radio-frequency cavities, spanning 2D geometry to 3D aerodynamics and electromagnetics, we show improved sample efficiency and the ability to explore beyond the confines of hand-crafted baselines.
The gas-phase chemistry of $\mathrm{C}_3\mathrm{H}_3^+$ plays a central role in the growth of carbonaceous species across a wide range of environments, from combustion systems to interstellar media. However, the presence of multiple isomers, primarily the linear propargyl cation and the cyclic cyclopropenyl cation, has so far prevented the determination of reliable isomer-specific reaction kinetics. In this work, we combine tunable vacuum-ultraviolet (VUV) photoionization with controlled ion--molecule reaction measurements to selectively produce and probe the reactivity of $\mathrm{C}_3\mathrm{H}_3^+$ isomers. By employing a titration approach, we extract isomer-specific reaction rate coefficients and branching ratios for reactions of $\mathrm{C}_3\mathrm{H}_3^+$ with $\mathrm{C}_2\mathrm{H}_4$, $\mathrm{C}_3\mathrm{H}_4$, and $\mathrm{C}_3\mathrm{H}_6$. Our results demonstrate a strong contrast in reactivity: the most stable cyclic isomer is found to be essentially unreactive under the investigated conditions, whereas the linear isomer exhibits efficient and diverse chemistry leading to the formation of larger hydrocarbon ions. In particular, the formation of $\mathrm{C}_6\mathrm{H}_5^+$ is observed, providing an important intermediate in pathways toward aromatic species. These findings help resolve long-standing ambiguities in previously reported rate coefficients and provide new quantitative constraints for kinetic and astrochemical models. More broadly, this work highlights the importance of isomer-specific chemistry in carbon growth processes and demonstrates the effectiveness of VUV-based photoionization methods for disentangling complex reaction networks in the gas phase.
Vision-language models (VLMs) show great potential for damage assessment after a disaster, but a recurring deficiency is that they are reluctant to declare a hazard; that is, recall is low even when overall accuracy appears adequate. This study examines that deficiency by using signal detection theory to decompose the decision behavior into perceptual capability and decision-criterion placement. We then propose a novel method for correcting the over-conservative decision policy, inspired by the finding that fear makes humans risk-averse, and ask whether an affective representation associated with fear can be causally manipulated to similarly alter a VLM's decision tendency. Using mechanistic interpretability, we localize a causally implicated affective circuit in the model and use activation steering to manipulate it while observing the effect on downstream prediction. The method is tested on a two-stage SeisMLLM pipeline built on Qwen2.5-VL-7B-Instruct, which flags only 27.0% of genuinely unsafe buildings on the SeisMLLM-1K test split and never issues a false Red, an SDT criterion of c = +1.354, despite adequate evidence quality (d' = 1.521). An affective direction is localized on emotion-rich natural scenes, causally validated by sparse-neuron knockout and distributed steering on held-out emotion data, and then injected into the building task. Fear-direction injection raises Red recall to 75.7% (p<0.001), and subtracting the same direction suppresses Red predictions entirely, whereas norm-matched random and matched happiness directions show no significant effect. The mechanism is a shift in criterion (c=-1.515) while discrimination is not improved (d'=-0.493). These results show that VLM decisions can be adjusted at inference time without retraining and demonstrate how mechanistic interpretability can be used to diagnose and control VLM decision behaviors in engineering applications.
Pinching-antenna (PA) systems provide spatial reconfigurability for integrated sensing and communication (ISAC) through controllable radiation locations along dielectric waveguides. This paper investigates a dual-waveguide PA-assisted ISAC system in which the communication user also serves as the sensing target, coupling the communication and sensing links through a common user/target geometry. Spatially separated transmit and receive waveguides provide separate spatial control over the forward and return sensing paths. We jointly optimize the transmit and receive PA positions and transmit power to maximize the aggregate communication rate subject to minimum sensing-SNR and two-dimensional localization Cramer-Rao bound (CRB) requirements, along with power, energy, and deployment constraints. To solve the resulting non-convex problem, we develop an alternating optimization algorithm that combines a bounded water-filling power update with a proximal first-order joint PA-position update. Numerical results show that the proposed design achieves higher communication rates than the considered benchmarks, with more pronounced gains under stringent localization requirements.