Searcharxiv⌕ Search

arXiv subjects

Search papers

Search indexed arXiv papers on artificial intelligence, large language models, computer vision and robotics. Read source abstracts and follow links to arXiv.

At least 379 records · Page 21Linked to original sources

Ramsey Theory for Product Trees

We develop a Ramsey theory for leaf-generated subsets of finite products of trees. Our starting point is a Theorem of Furstenberg and Weiss which states that for every $k\geq 1$ and $α>0$, if $A$ is a subset of the leaves of $T_n$, where $T_n$ is the complete binary tree of height $n$, with size $|A|\geq 2^{αn}$, then for $n$ sufficiently large the ancestor closed sub-tree $T_A\subset T_n$ generated by $A$ must contain a copy of $T_k$ such that (1) all vertices in the same level of $T_k$ are mapped into vertices at the same level of $T_A$, (2) if a non-leaf vertex $x\in T_k$ is mapped into a vertex $y \in T_A$, then the two children of $x$ are mapped into descendants of the two children of $y$, and (3) the levels of $T_A$ occupied by the copy of $T_k$ form an arithmetic progression. For the product $T_n \times T_n$ of two binary trees, we show that for every $k\geq 1$ and $α> 1$, for $n$ sufficiently large, every subset $A$ of the leaves of $T_n\times T_n$ of size at least $2^{αn}$ has that its ancestor closure $Γ_A\subset T_n \times T_n$ contains similarly structured arithmetic copies of the height $k$ four-ary tree, where the four children of every non-leaf vertex in the source tree are required to map below the four distinct \textit{diagonal children} of the image vertex, where a diagonal child of a vertex $(x,y) \in T_n \times T_n$ is obtained by moving one level down each in each component. Our results generalise to product of $d$-many finite $b$-ary trees with a sharp critical exponent of \[ α_{\text{crit}}(d,b) =d-1 + \log_b(b-1).\] Geometrically, our results imply that any set $E\subset [0,1)^d$ of upper Minkowski dimension greater than $α_{\text{crit}}(d,b)$ contains arithmetic $b$-adic branching patterns of arbitrary finite order.

math.CO↗

When Do Differentially Private Inputs Protect Graph Shift Operators?

We study the differential privacy (DP) of a graph shift operator (GSO) when an analyst observes the output of a graph filter. In particular, we study the setting in which the input signals to the graph filter are drawn from a differentially private distribution. Unlike approaches that perturb the GSO or the filter output, we use the randomness already present in the inputs to protect the GSO. This yields an equivalent level of privacy protection to that of the perturbation methods without adding noise, and thus a better privacy-utility trade-off. We provide an explicit characterization of the privacy loss and its certificate in terms of the zeros of the graph filter. In doing so, we show that the log-likelihood ratio between the releases of two adjacent topologies is governed by the distances from each zero to the graph frequencies of the two GSOs. Then, by uniformly bounding the log-likelihood ratio over the adjacent topologies, we obtain an explicit $(\varepsilon,δ)$-DP guarantee for Gaussian inputs. We further show, via a Cramér--Rao bound, that the zero placement that limits the privacy loss also raises the floor on the adversary's reconstruction error. Finally, empirical validation is performed on a synthetic network of financial exposures, where the largest position a pair can conceal and the accuracy with which it can be sized are collinear across pairs. Both are set by the graph-frequency content of the pair, and the full network becomes recoverable only as the certified budget grows.

cs.CR↗

Codetta: High-Capacity, Keyless, and Undetectable Multi-Agent Collusion

Multi-agent systems built on large language models (LLMs) are increasingly deployed in high-stakes settings such as finance, healthcare, and software engineering, where agents coordinate through natural-language messages. The same channels, however, let colluding agents exfiltrate confidential information or coordinate unauthorized actions, and steganography can hide such communication inside outputs that look ordinary to an auditor reading the transcript. Existing provably undetectable LLM steganography protocols are not suited to realistic deployments. High-capacity schemes assume a symmetric setting where the receiver can reproduce the sender's output distribution, the state-of-the-art protocol for asymmetric agents has very low capacity, and most approaches rely on a pre-shared secret key. We make the threat of undetectable agent collusion concrete with Codetta, a high-capacity steganographic protocol for independently deployed agents in realistic asymmetric settings. Codetta combines a shared public model that estimates the communication channel, a sampling mechanism that preserves the sender's output distribution, and an adaptive error-correcting code. It further removes the pre-shared key through a steganographic key exchange that lets independently deployed agents establish a shared key while keeping the transcript computationally indistinguishable from ordinary model outputs. Across three agent workloads and three sender models, Codetta achieves up to $94\times$ the capacity of the state-of-the-art asymmetric protocol, and its key exchange establishes a shared key with about 80k visible tokens at an empirically certified failure probability of at most $4.1\times 10^{-3}$. These results show that effectively undetectable collusion is becoming feasible between independently deployed agents, so auditing must go beyond inspecting communication transcripts.

cs.CR↗

A Quantum Circuit for Gaussian Elimination

A quantum circuit for Gaussian elimination is developed in this work. While previous quantum implementations of Gaussian elimination were restricted to $\mathrm{GF}(2)$, the proposed design is generalized to any finite field. Compared with previous works in $\mathrm{GF}(2)$, the developed circuit achieves garbage-free construction while retaining the best known asymptotic Toffoli depth up to logarithmic factors.

quant-ph↗

On Fast-Slow Mean-Field Forward-Backward Stochastic Systems

We establish an averaging principle for a class of multiscale mean-field forward-backward stochastic differential equations and identify several novel phenomena that are absent from classical fast-slow systems. In contrast with classical fast-slow systems, the effective dynamics cannot in general be obtained by simply freezing deterministic slow parameters and averaging against the invariant measure of the resulting fast equation. The appropriate averaging object is instead provided by a frozen fast dynamics in a random environment and its associated conditional invariant measures, which retain the coupling between the slow state and its distribution. The forward-backward structure creates a further obstruction: local averaging estimates need not remain stable when propagated over an arbitrary time horizon. We identify a uniform restart stability condition for the averaged system under which this obstruction can be overcome. Using a joint lifted semigroup for the state-law dynamics, together with a two-scale discretization and a Gordin-type decomposition, we prove strong averaging for both the forward and backward components with optimal convergence rate $O(\varepsilon^{1/2})$. As an application, we apply the general theory to a class of mean-field stochastic control problems and develop an efficient algorithm for solving such mean-field control problems.

math.OC↗

Conductance of silicon nanotube junctions in high magnetic fields

We investigate coherent quantum transport through silicon nanotube (SiNT) junctions in high magnetic fields up to 60 T using a tight-binding model combined with the non-equilibrium Green's function formalism, and magnetic field included via Peierls substitution. We consider junctions of metallic nanotubes (6,0)+(6,0) and semiconducting ones (9,9)+(9,9), and examine the effects of the overlap length, inter-tube distance, magnetic-field direction, and field strength on the electronic transmission. In contrast to carbon nanotube junctions, the SiNT systems exhibit irregular transmission oscillations and do not show the emergence of highly conductive gateway states. The transmission is substantially more sensitive to a magnetic field perpendicular to the nanotube axis than to a parallel field, while increasing the field strength progressively modifies the transmission spectrum. Increasing the overlap length results in more frequent transmission oscillations, whereas increasing the inter-tube distance modifies their positions and amplitudes without changing their overall character. Generally, similar trends are observed for both types of junctions, involving metallic and semiconducting nanotubes. However, in the junction of semiconducting nanotubes, one observes peculiar additional field-dependent in-gap transmission features. These results demonstrate that the magnetic-field response of SiNT junctions is strongly governed by their geometry and differs qualitatively from that of pristine carbon nanotube junctions.

cond-mat.mes-hall↗

Implementation and Evaluation of NTT Arithmetic for ML-KEM on a CGLA

FIPS 203 standardizes ML-KEM for post-quantum key establishment. Its polynomial multiplication relies on NTT butterflies with exact modular arithmetic over q = 3329. Dedicated NTT accelerators minimize latency with fixed modular arithmetic and stage schedules. A CPU-Grounded Linear Array (CGLA) reuses one programmable linear datapath across several workloads. Mapping the transform to this datapath requires exact FP32 reconstruction and explicit stage transitions across the ARM-to-CGLA interface. We implement an eight-stage cyclic radix-2 driver over the ML-KEM modulus by splitting each twiddle into 8-bit and 4-bit parts before modular reduction. This driver differs from the standardized seven-layer incomplete negacyclic NTT and does not implement the full ML-KEM polynomial multiplication path. The arithmetic sequence keeps every integer below 2^24. One 41-PE call fuses the first two radix-2 stages, and six 47-PE calls execute the remaining stages while scattering outputs into next-stage records. Across four cohorts, 105 FPGA runs match all 817152 output coefficients. At batch size 64, measured FPGA end-to-end latency is 27.5 us per NTT, and the ASIC projection is 6.74 us. With PE gating, projected ASIC system energy is 10.1 uJ per NTT at batch size 8 and 58.9 uJ at batch size 64.

cs.AR↗

On the degenerate Weyl problem on isometric immersions

This paper is concerned with the Weyl problem, \emph{i.e.}, the existence of isometric immersions or embeddings of two-spheres into the three-dimensional Euclidean space or general ambient three-manifolds. We establish two results on the degenerate Weyl problem, namely when the Gaussian curvature $K_g$ is only nonnegative rather than strictly positive. First, for a smooth metric $g$ on the two-sphere, if $K_g$ is strictly positive except at finitely many points where the Hessian of $K_g$ is positive definite, then $g$ admits a global $C^{2,1}$-isometric embedding into $\mathbb{R}^3$. It appears to be the first result on the degenerate Weyl problem with purely intrinsic conditions on $g$. Second, for a general simply-connected ambient three-manifold $(\mathcal{M},{\overline{g}})$, if $K_g \geq K_0 \geq {\rm sec}_{\overline{g}}$ for some constant $K_0$, $(K_g-K_0)^{-1/2} \in L^p$ with $p \geq 2$, and a certain uniform pinching condition holds for approximate nondegenerate isometric immersions, then there exists a $W^{3,p}$-isometric immersion. Alongside we also resolve the nondegenerate Weyl problem (\emph{i.e.}, when $K_g>0$) into general simply-connected ambient three-manifolds for $g$, ${\overline{g}} \in C^{2,1}$.

math.DG↗

ReaFlow-TTS: Realization-Conditioned Flow Matching for High-Quality and Controllable Speech Synthesis

In flow-matching text-to-speech (TTS), different speech realizations can induce different target velocities under the same generation conditions. A deterministic velocity field trained with squared error predicts their conditional mean, thereby marginalizing realization-dependent variation. Meanwhile, modeling such variation does not inherently provide a semantically interpretable interface for attribute manipulation. We propose ReaFlow-TTS, a realization-conditioned flow-matching framework that introduces an utterance-level stochastic realization latent and uses it to condition velocity prediction throughout the generation trajectory. We further impose valence-arousal-dominance (VAD) semantics on the realization space, enabling direct and graded attribute manipulation without target speech at inference. Experiments demonstrate improved synthesis quality over a matched full-mask baseline and reproducible latent-induced pitch, energy, and timing tendencies across initial-noise samples, providing behavioral evidence that the latent is used as a reusable realization condition. Subjective evaluation further demonstrates graded VAD manipulation across generation contexts with only modest changes in naturalness.

cs.SD↗

On symplectic aspects of $SU(2)$ character varieties for punctured surfaces

For a surface with an odd number of punctures, the moduli space of flat $SU(2)$ connections with traceless holonomy around each puncture is a symplectic manifold. When the moduli space is nonempty, there is a natural homomorphism from the mapping class group of the punctured surface to the symplectic mapping class group of this moduli space. It is shown that this homomorphism is injective if and only if the dimension of the moduli space is greater than $2$. This generalizes work of Seidel and Wehrheim--Woodward. Also given is a complete classification of Lagrangian spheres in the projective plane blown up at $5$ points with its monotone symplectic structure, which is the moduli space for the 5-punctured sphere. Furthermore, it is determined when two such Lagrangian spheres can be displaced by a symplectic isotopy. Results are also obtained regarding Lagrangian spheres in the intersection of two quadrics in $\mathbb{C}\mathbb{P}^5$. The proofs involve instanton Floer theory and results on Heegaard splittings. A main technical result establishes the approximation of any Hamiltonian isotopy of the $SU(2)$ moduli space by holonomy perturbations which are used in instanton homology.

math.GT↗

Automatic Harness Evolution for Hardware Design Verification: Can LLMs Consolidate Gains Across Discovered Harnesses?

Agent behavior depends on the harness surrounding a language model, but it remains unclear whether language models can reliably improve such harnesses for hardware-design tasks. We study automatic harness evolution around a fixed subject model on 12 proprietary design-verification root-cause localization tasks. Across five trials per task, automatically evolved harnesses increased completed attempts by 71-76% and any-hit task coverage by 80-100%, while total correct attempts improved by only 18-24%. The strongest success reproducible at least twice result improved by one task, and later candidates exchanged gains across tasks rather than preserving them. An auxiliary candidate improved on a four-task validation set excluded from search but tied its baseline on a subsequent 12-task replay containing both search and validation tasks, so the selected gain did not persist across the full pool. Across the tested lineage, useful search, evidence, and finalization behaviors appeared in different candidates but did not consistently consolidate into a single harness that dominated across tasks and metrics. In a separate CVDP cross-benchmark case study, an automatically evolved defined-width repair harness produced 35.6% more functional passes than its 142-task reference baseline; the final functional verifier scored completed outputs but was not shown to the subject agent during repair. These results support archive-aware selection when evolution yields complementary specializations without consistent consolidation.

cs.SE↗

Squaring Up by Selection: NP-Completeness at Three Simple Roots

To solve an overdetermined polynomial system numerically, one first makes it square, usually by replacing the given equations with as many random linear combinations as there are unknowns. This is a provably safe step, but it can substantially enlarge the supports. The alternative is to keep that many of the given equations themselves. Selection preserves sparsity but risks geometry: a genuine solution can cease to be an isolated point of the subsystem's zero set. We show that deciding whether a safe choice exists is NP-complete, already for an explicit family of systems of degree three with radical ideal and exactly three simple rational solutions. For strong selection with the nondegenerate rational solutions supplied explicitly, three is the exact threshold when degrees are polynomially bounded: one or two solutions reduce to matroid intersection, three already give NP-completeness. Even without a degree bound, an arbitrarily long list never takes the decision problem beyond NP. On the hard family, five natural notions of a faithful subsystem coincide, and every failing choice fails visibly: its zero set contains an affine subspace through one of the three solutions. A degree-four variant shows that cost information does not help: every candidate that could possibly succeed has mixed volume exactly three, and the problem is NP-complete still. The construction realizes Karp's three-dimensional matching problem as the selection of a square subsystem from the given equations.

math.AG↗

Robots That Take Initiative: A Framework for Building and Evaluating Proactive Robots

Effective robot assistance beyond narrow roles and repetitive tasks requires robots to be proactive - to decide what needs to be done rather than waiting to be told. While proactivity is increasingly explored, it lacks a unified formulation, and work in the domain is typically evaluated offline against static human models that cannot capture the effect of a robot's actions on the environment and the user's own behavior. We introduce a unified formalism for proactive robot assistance, organize it into three levels, and provide a framework to address the highest level of unprompted proactive assistance. We then show that offline evaluation overstates performance in this setting, and contribute a closed-loop evaluation with a human model that adapts to the robot. Finally, we present a method, GAP, that instantiates our framework, learning from passive observation to anticipate user goals and act. Under closed-loop evaluation, prior state-of-the-art methods collapse, in some cases adding more work than they save, while GAP remains robust and substantially outperforms them.

cs.RO↗

LoRa Fluid Antenna Multiple Access

Concurrent long-range (LoRa) transmissions over the same time-frequency and spreading factor (SF) resources generally result in packet collisions, as the gateway cannot distinguish the overlapping signals from different end devices (EDs). This paper advocates a new fluid antenna multiple access (FAMA) framework for LoRa, referred to as {\it lora}-FAMA, to provide spatial opportunities for LoRa multiuser communications. In {\it lora}-FAMA, a gateway employs a single fluid antenna connected to only one radio-frequency (RF) chain, whose radiating element traverses the antenna aperture by sequentially visiting all candidate positions, i.e., fluid antenna `ports', within each symbol interval. The signal segments collected along the trajectory are compensated using the channel state information for the target ED. As a result, the desired signal is coherently accumulated, whereas the signals from other EDs experience unmatched channel variations and thus cannot be coherently combined. Applying the compensation separately to each active ED enables simultaneous multiuser transmission. We analyze the statistical performance of {\it lora}-FAMA under asynchronous transmissions and spatially correlated fading, as well as the large-aperture limiting case with independent and identically distributed fading. Numerical results show close agreement between the analytical and Monte Carlo results. It is revealed that, with a normalized aperture of $4\times4$ and $\mathrm{SF}=9$, a gateway can simultaneously serve more than $10$ EDs over the same frequency and SF resources while maintaining a symbol error rate below $10^{-4}$. These results demonstrate the potential of fluid antennas to enable LoRa multiple access without multiple RF chains.

eess.SP↗

Dense pentacene cocrystal demonstrates room-temperature coherent control

Maximizing the number of addressable spins within a fixed volume can improve ensemble quantum sensor sensitivity, but dense packing usually increases dipolar interactions and excited-state transport, shortening coherence and suppressing optical readout. Here we report a 2:1 cocrystal of 6,13-dihydropentacene and pentacene (DHP/Pc) containing 33.3 mol% pentacene ($3.3\times10^{5}$ ppm; $9.51\times10^{20}$ cm$^{-3}$), a volumetric spin-site density more than two orders of magnitude above previous benchmarks, NV-diamond and pentacene-doped p-terphenyl (PDP). Despite this density, DHP/Pc exhibits microsecond spin coherence, room-temperature optically detected magnetic resonance and coherent control. Time-resolved measurements indicate that the long-lived triplet population is generated predominantly by intersystem crossing and exhibits strong, non-thermal sublevel polarization. Density-functional theory calculations further suggest that the native cocrystal geometry weakens electronic coupling between neighboring pentacenes, while the higher DHP triplet energy creates a barrier to triplet migration. These results identify molecular packing and coformer triplet energetics as complementary design parameters for preserving coherence in high spin-site density systems, demonstrating cocrystallization as a promising route to dense, optically addressable spin materials for room-temperature ensemble quantum sensing.

quant-ph↗

Moiré droplet of ultracold Bose gases in a twisted-bilayer optical lattice

We report the emergence of Moiré droplet in two-dimensional ultracold bosons subjected to a spin-dependent optical lattice, effctively realizing a twisted-bilayer configuration. We show that the droplet formation dramatically enhances the visibility of Moiré pattern in the density profile, even for exceptionally weak lattice potentials. The Moiré pattern can be enhanced similarly by increasing the lattice depth, which, however, also induces droplet diffusion characterized by a spreading density profile. Furthermore, we demonstrate a dynamical generation of Moiré pattern by dragging a small droplet through a moving lattice. At appropriate velocities, the droplet undergoes bifurcation and exhibits pronounced Moiré pattern within periodic time intervals. Our results establish the ultracold droplet as a compelling platform for simulating interacting Moiré physics, particularly the interplay between Moiré lattice and bound-state formation.

cond-mat.quant-gas↗

Turning-Point Count Discrepancy as a Diagnostic of Relativistic Orbital Chaos

We propose the turning-point count-discrepancy indicator (TPCD) for diagnosing orbital chaos from a single trajectory in relativistic Hamiltonian systems with two oscillatory degrees of freedom. TPCD measures the largest cumulative departure of one turning-event count from its mean rate per reference cycle, requiring neither a neighboring orbit nor a phase-space partition and applying to both massive particles and photons. We establish its long-time behavior under explicit event--phase assumptions. Rigid phases with an exact event--phase correspondence obey a strict discrepancy bound of unity, and linearizable regular tori with bounded degree-one phase deformations obey a finite, orbit-dependent bound; both imply that the normalized indicator decays to zero as the record grows. A diffusive fluctuation mechanism instead yields a Brownian-bridge scaling and a finite statistical scale. Integrable Kerr motion validates the construction, recovering prescribed frequency ratios from event counts to within $2.6\times10^{-5}$ for six targets, including an irrational ratio. In charged-particle scans around a Kerr black hole in an external test magnetic field, TPCD and the fast Lyapunov indicator agree for all 80 sampled trajectories. In the Schwarzschild--Melvin photon model, a trajectory with elevated finite-time TPCD but low fast-Lyapunov values is identified as regular once its indicator trends downward over an extended integration, showing that finite-time values must be read together with their long-time trend.

gr-qc↗

On the Effectiveness of Kernel-Level Evidence for Agent Security

LLM agents are deployed into infrastructure that grants them broad host authority, yet existing agent-security benchmarks and defenses operate almost exclusively at the application telemetry layer: the served tool manifest, the user prompt, and the model's messages. Some threats, however, smuggle malicious instructions and actions past the application boundary, leaving them invisible to that layer. In this work, we bridge that gap by pairing application-level agent telemetry with kernel-level syscall traces to present the first paired-evidence characterization of kernel-level versus application-layer signal for agent security. To quantify the value of the enhanced telemetry, we introduce Agent Cross-Layer Evidence (ACE), a paired-session corpus of 4,047 sessions and 17 threat models spanning six delivery-vector families and 14 of the 25 OWASP LLM and agentic threat categories, organized into 12 attack mechanics with per-mechanic characterization of where the most discriminative evidence lies. Across four distinct detector families, we find that kernel evidence is discriminative on its own and that composing it with application-layer evidence generally outperforms either single-layer view, revealing complementary signals that single-layer analyses can miss. We further demonstrate generalization to unseen attack families and transfer to an alternate agent runtime. Together, these findings establish the value of cross-layer evidence for agent security.

cs.CR↗