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 451 records · Page 25Linked to original sources

arXiv2026

Communication-Efficient Relative Pose Estimation with Vision Foundation Models for Ephemeral Collaborative Perception

Relative pose estimation is a fundamental capability for collaborative perception and coordination in multi-robot systems. However, robots encountering each other in real-world environments often operate in short interaction windows and must operate under limited communication bandwidth with intermittent or missing visual overlap caused by occlusions or limited fields of view. Existing approaches typically rely on global reference frames, assume sustained view overlap, or incur prohibitive communication costs, thereby limiting their applicability to ephemeral collaborative perception. To address these challenges, we introduce communication-efficient relative pose estimation (CERPE), a system-level framework that coordinates vision foundation models to jointly estimate ego-motion and inter-robot relative pose. CERPE reduces unnecessary raw-observation exchange by using continuously shared fixed-size descriptors to gate event-triggered raw-image requests independently of pose estimation. Non-overlapping encounters are handled by propagating inter-robot relative poses through metrically scaled ego-motion, thus maintaining relative pose estimates even in the absence of visual overlap. Experiments in simulation and real-world robots show that CERPE improves 6-DoF relative pose estimation over selected baselines in ephemeral collaborative perception.

cs.RO↗

arXiv2026

Rubrics on Trial: Evolving Rubrics from a Single Query via Synthetic Pairwise Evidence

Rubric evolution offers a promising approach to improving the quality of rubrics generated by large language models (LLMs). Central to this process is rubric comparison, which identifies the better of two rubrics and guides the direction of evolution. However, accurate rubric comparison is difficult, which presents two challenges. (1) It should reflect downstream task performance, which is essential for assessing rubric utility but often prohibitively expensive to evaluate. (2) It should discourage unnecessary criteria, which increase verification costs and may dilute the influence of essential criteria. To address these challenges, we introduce Rubrics on Trial, a multi-agent framework that evolves rubrics by comparing synthetic response pairs. To address challenge 1, the framework compares synthetic responses that satisfy the respective rubrics, providing a proxy for downstream performance without training a separate policy for each rubric. To address challenge 2, it assesses the necessity of a candidate criterion by independently generating high-quality alternative responses that violate it and comparing them with edited versions that satisfy it. A rubric is favored when it improves response quality in both comparisons, and the resulting comparison signal is further incorporated for rubric evolution. Extensive experiments demonstrate that Rubrics on Trial improves the quality of generated rubrics and leads to better downstream task performance.

cs.CL↗

arXiv2026

Spectral study of the outburst decay of the accreting millisecond X-ray pulsar SRGA J144459.2-604207

We study the spectra of the accreting X-ray millisecond pulsar SRGA J144459.2-604207 during the 2024 outburst using the Neutron star Interior Composition Explorer (NICER) and Swift observations. The spectra during the outburst decay, reflares, and quiescent state are explored using the absorbed Comptonized model. We find that the spectra during the quiescent state can also be explained using the absorbed power-law and absorbed blackbody model. The spectral evolution of the source is explored as the outburst decays into quiescence. We study the long-term quiescent X-ray activity of the source spanning roughly 45 years and find that the long-term quiescent luminosity may be explained using the deep crustal heating model. We also find that the coronal activity of the companion star alone cannot power the quiescent X-ray luminosity of the source. We place the source on the radio-X-ray luminosity plane and compare its position with other sources. We estimate the propeller luminosity of the source and find that it is smaller than the estimated luminosity during reflares and the quiescent state during the 2024 outburst. Several reflares are detected during the outburst decay, one of which is near-simultaneous with the detection of an ultrafast outflow and radio emission. We explore plausible mechanisms that may power outflow, radio emission and associated jet formation in this accreting binary.

astro-ph.HE↗

arXiv2026

Counterexamples to additivity of minimum output $p$-Rényi entropy of quantum channels for all $p\ge 0$

The additivity of minimum output entropies is a central problem in quantum information theory. Nonadditivity is known for every Rényi order $p>1$, at the von Neumann point $p=1$, and for sufficiently small positive $p$, while much of the interval $0 0$ there exist finite-dimensional quantum channels whose minimum output $p$-Rényi entropies violate additivity. Our proof combines two constructions: random projection-induced channels yield nonadditivity for $p>3/4$, and antisymmetric postprocessing extends the violation to all positive Rényi orders. Our estimates also improve the output-dimension bound obtained by Belinschi, Collins, and Nechita for additivity violation of minimum output von Neumann entropy.

quant-ph↗

arXiv2026

Cura 1T: Healthcare Foundation Model via Recursive Self-Improvement

Healthcare spans high-stakes communication, expert reasoning, and workflow execution, yet specialized language models that cover these use cases together remain limited. A healthcare model must handle patient consultation, clinical reasoning over text and images, interactive diagnosis, and electronic health record (EHR) tool use. These capabilities fail in different ways, and a narrow update for one task can degrade another. We present Cura 1T, a healthcare foundation model trained through recursive self-improvement (RSI). In each RSI round, the RSI harness runs the current model on healthcare benchmarks, evaluates the trajectories to locate capability gaps, and refines the training mixture by synthesizing training data. On 6 healthcare benchmarks, Cura 1T scores highest on MedAgentBench, HealthBench Professional, HealthBench Hard, MedXpertQA text, and AgentClinic, and second on MedXpertQA multimodal. It preserves performances on out-of-domain reasoning and agentic benchmarks including AIME, GPQA-Diamond, and $τ^2$-Bench.

cs.AI↗

arXiv2026

Solving Stackelberg Vertex Cover on trees using split and join

The Stackelberg Vertex Cover problem is a bilevel optimization problem with two players on a graph G = ($F \cup P$, E) where each vertex from F has a weight and the first player selects a price for each vertex in P . Afterwards, the second player finds a minimum weight vertex cover X and the first player receives the set price for each vertex from $X \cap P$ . The goal is to maximize the revenue of the first player. This problem was recently shown to be NP-complete for bipartite graphs while being solvable in linear time on paths. We present four new algorithms for solving Stackelberg Vertex Cover on certain kinds of graphs: (1) a pseudo-polynomial algorithm working on general trees when all weights are integer with a runtime linear in the number of vertices and cubic in the maximum weight (2) a generalization of (1) for bipartite graphs with integer weights and a tree decomposition that is FPT in the maximum weight and the treewidth, (3) a strongly polynomial algorithm for rooted trees having the property that the least common ancestor of any two vertices from P is again in P (this case includes paths); and (4) an FPT-algorithm for trees, where the parameter is the maximum number P-vertices $v_i$ that an F-vertex u can reach while using no other P -vertices. These algorithms are based on a lemma that allows us to split instances at a vertex u into multiple sub-instances, which follows from LP duality and integrality of the vertex cover LP on bipartite graphs. The lemma requires that the minimum vertex covers of the sub-instances agree on u (either all include u or all don't). For this we introduce the concept of commitments. We show that the Stackelberg Vertex Cover problem with commitments is weakly NP-complete. An open question is the non-bipartite case as there is an explicit counterexample showing that the split-and-join technique does not work.

cs.DS↗

arXiv2026

Hawking evaporation and the possibility of remnant formation for covariant quantum black holes

Covariant quantum black holes (BHs) are an important development in loop quantum gravity (LQG), and investigations of their evaporation may offer a possible avenue for testing LQG. In this paper, we investigate the evaporation of covariant quantum BHs and the possibility of remnant formation. Our results show that covariant quantum BHs can leave remnants, calculated by analytical greybody factors for a massless scalar field. Nevertheless, the fully numerical results indicate mass loss rates of covariant quantum BHs are spin-dependent and differ from those of Schwarzschild BHs, due to spin-dependent greybody factors. In particular, these BHs evaporate faster than Schwarzschild BHs in the low-mass regime, with the evaporation being dominated by massless neutrino emission rather than scalar field emission, contrary to the scalar field dominance commonly assumed in previous studies. Therefore, considering only BH evaporation in the massless scalar field case may be insufficient and may even lead to misjudging BH fate. In addition, these results may provide a way to test LQG in the future.

gr-qc↗

arXiv2026

A Dual-Hypothesis Reasoning Framework for LLM Guardrails

We propose ARBITER, a novel LLM guardrail framework that introduces two key ideas: (i) dual-hypothesis reasoning, a reasoning method for LLM guardrails that explicitly considers both safe and unsafe interpretations of a prompt before making a safety decision, and (ii) multi-component supervised fine-tuning (MC-SFT), a structured training loss for reasoning-based guardrails that decomposes LLM outputs into logical components and weights them according to their importance. Existing reasoning-based guardrails often rely on expensive procedures, such as generating reasoning traces using larger or closed-source teacher models and applying full-parameter fine-tuning. In contrast, ARBITER uses a cost-effective self-generation strategy for reasoning traces and LoRA-based parameter-efficient fine-tuning while still achieving better performance than these expensive approaches. Additionally, ARBITER provides faithful evidence-phrase explanations for unsafe decisions, enabling a more transparent and interpretable guardrail method. Experiments on three safety moderation benchmarks show that ARBITER outperforms existing reasoning-based and non-reasoning guardrail baselines, with clear gains in out-of-domain evaluations.

cs.AI↗

arXiv2026

The Label Complexity of Useful Class-Conditional Prediction Sets under Distribution Shift

Prediction sets can make deployed classifiers safer by returning several plausible labels when a single prediction is uncertain. Their value depends on classwise reliability: average coverage can meet its target while rare or difficult classes fail repeatedly. This concern is sharper after distribution shift, when calibration labels come from a source environment but reliability is needed on the target. We ask what labeled source data and unlabeled target inputs reveal about class-conditional prediction sets, and when target labels are necessary. Under unrestricted joint shift, two target laws can produce the same observable data while requiring different classwise thresholds; any label-free rule covering both must enlarge its sets on one law. We give a labeled target audit that estimates the missing quantiles with a simultaneous guarantee. Probability-scale error is invariant to increasing score transformations, and threshold recovery follows under local regularity. At fixed confidence, achieving threshold tolerance $\varepsilon$ with fixed, nonadaptive class-stratified labeled pairs has total complexity $Θ(K\varepsilon^{-2}\log K)$, or $Θ(\varepsilon^{-2}\log K)$ labels per class under equal allocation. Class imbalance creates a separate acquisition cost; for foreground class probabilities of order $1/K$, the mixed-stream label complexity is also $Θ(K\varepsilon^{-2}\log K)$ at fixed confidence. Experiments on action-recognition and image shifts show that marginal coverage can conceal severe class failures and that source classwise calibration depends on the shift. The results connect the information available at deployment to the target labels needed for useful class-conditional prediction.

cs.LG↗

arXiv2026

The Dynamics of Discoverability: How Trajectories and Priors Shape Equation Recovery

How the dynamical regime of the observed system affects equation discovery has mainly been investigated through comparisons across systems. However, such comparisons vary both the equations and the dynamics, confounding the effect of the regime with the difficulty of recovering the equations symbolically. We separate the two by varying the forcing of Lorenz-84, moving its fully observed post-transient trajectories through fixed-point, periodic, and chaotic regimes while preserving the equations' functional form. Within each regime, we separately vary the amount of data, the noise, and prior knowledge of which terms the equations contain. We then measure how well two complementary approaches, sparse regression over a fixed library of candidate functions (SINDy) and an evolutionary search over symbolic expression trees (PySR), recover the true equations' terms and coefficients. We find that recovery depends on whether the sampled states distinguish combinations of candidate functions: equations remain poorly recovered from fixed-point data even when the candidate set contains only the true terms. We link the effect of the dynamics on both algorithms to one object: the moment matrix of the candidate functions under the invariant measure of the regime. Small eigenvalues mark weakly distinguishable combinations of candidate functions: we show that more data, less noise, and more prior knowledge can mitigate the resulting recovery difficulties, while a zero eigenvalue makes distinct equations indistinguishable on the visited states. Hence, a more precise prior needs less informative data, with consequences for data collection, method design, and evaluation.

cs.LG↗

arXiv2026

On a certain arithmetic function defined via Bernoulli numbers

The paper defines a certain function $χ(n)$ on the set of odd integers $n \ge 3$. The integrality property of this function is investigated. The main result establishes that the function is integer-valued if and only if $n$ is a prime number or an odd Giuga number. Thus, the integrality of $χ(n)$ serves as a single criterion unifying three classes: prime numbers, Carmichael numbers, and odd Giuga numbers. The proof relies exclusively on the von Staudt--Clausen theorem on the denominators of Bernoulli numbers and Korselt's criterion for Carmichael numbers, which makes the exposition elementary. A connection is proven between prime numbers $p$ for which $χ(p)=p-1$ and Novák--Carmichael numbers. It is proven that the primes satisfying this criterion form a sequence with the identifier A337119 in the Online Encyclopedia of Integer Sequences. For the classes $P_μ$ of primes with a fixed value $χ(p)=μ$, a characterization is obtained in terms of the exponents of the prime divisors of the numerator of the corresponding Bernoulli number $B_{p-1}$. For some $μ$, explicit congruences are provided. The work is accessible to a wide readership due to its minimal use of advanced mathematical apparatus.

math.NT↗

arXiv2026

Electrostatic Control Enables Robust Helical Edge Channel Transport in III-V Quantum Spin Hall Insulators

Quantum spin Hall transport in InAs/GaInSb-based two-dimensional topological insulators can be limited by parasitic bulk and edge contributions. We demonstrate that these limitations are effectively mitigated through electrostatic control in dual-gated InAs/GaInSb/InAs trilayer quantum wells grown on AlSb quasi-substrates. In macroscopic Hall bars exceeding the phase coherence length, a multi-probe analysis reveals an insulating bulk and a constant edge resistance over a wide electric-field range. In microscopic devices with edge lengths below the phase coherence lengths, the edge resistance remains robust and quantized across a broad field range, revealing the intrinsic resilience of helical edge channels to electric-field perturbations. Only beyond a threshold value, parasitic edge contributions emerge. These results establish dual gating as a reliable strategy to suppress parasitic conduction while stabilizing helical edge transport, providing a versatile and reproducible platform for tunable topological transport in III-V quantum spin Hall systems.

cond-mat.mes-hall↗

arXiv2026

In-Context Time Series Classification with Random Convolutional Features

Time series classification is central to domains such as medical signal analysis, industrial monitoring, and sensor-based activity recognition, where class information manifests as localized shapes, specific frequencies, temporal shifts, or complex cross-channel interactions. Random convolutional transforms capture these diverse patterns by converting time series into rich, fixed-dimensional feature representations that can be processed by standard tabular classifiers. While these representations are traditionally paired with simple linear models, we investigate whether a pretrained tabular foundation model can exploit them more effectively and how its performance depends on the available data and inference budget. We propose MASHT, a pipeline that combines MultiRocket and Hydra features with an in-context tabular foundation model. Our approach uses a pretrained tabular foundation model to bypass task-specific model training, requiring only feature extraction and direct inference. Extensive experiments demonstrate that MASHT matches state-of-the-art time series classification baselines on univariate tasks, achieving a lower average rank than HIVE-COTE 2.0. On multivariate datasets, MASHT remains highly competitive with the strongest reference methods. Controlled resource experiments show that compact feature tables retain most of the accuracy at substantially lower runtime, while TabPFN outperforms a matched linear baseline across the evaluated label budgets on univariate tasks. These results highlight practical trade-offs between predictive performance, labeled data, and inference cost.

cs.LG↗

arXiv2026

Measurement of the $γγ\to ττ$ cross section and constraints on the anomalous magnetic moment of the $τ$ lepton in ultraperipheral PbPb collisions at $\sqrt{s_\mathrm{NN}}$ = 5.02 TeV

The production of $τ$ lepton pairs via photon-photon fusion, $γγ\to ττ$, is studied in ultraperipheral lead-lead collisions at a nucleon-nucleon center-of-mass energy of 5.02 TeV. The dataset, collected by the CMS experiment in 2018, corresponds to an integrated luminosity of 1.70 nb$^{-1}$. Four different $τ^+τ^-$ decay final states are analyzed. A simultaneous likelihood fit to the measured lepton transverse momentum ($p_\mathrm{T}$) distributions, which incorporates information from both spectral shape and normalization, is used to constrain the anomalous magnetic moment of the $τ$ lepton, $a_τ$, and to extract the cross section of the process. The measured 95% CL interval for $a_τ$ is $-$0.039 $\lt$ $a_τ$ $\lt$ 0.032. The fiducial cross section, $σ_{γγ\to ττ}^\text{fid}$ = 555$^{+60}_{-14}$ $μ$b for tau leptons with $p_\mathrm{T}^τ$ $\gt$ 1 GeV and pseudorapidity $\lvertη^τ\rvert$ $\lt$ 3, is the most precise measurement for this process at the LHC to date and is in agreement with next-to-leading-order quantum electrodynamics predictions.

nucl-ex↗

arXiv2026

Collisionless stationary states of a stratified plasma in an expanding magnetic tube with stochastic heating

We investigate the collisionless kinetic structure of the upper solar atmosphere in an expanding magnetic field. Building on established velocity-filtration models, we extend the stochastic multi-temperature framework developed in previous works by including magnetic-field expansion and magnetic-moment conservation. Stochastic heating generates a non-Maxwellian boundary distribution through the superposition of particle populations associated with different temperatures. We consider a stationary two-component plasma confined within an expanding magnetic flux tube and subject to gravity, self-consistent electrostatic interactions, the Pannekoek--Rosseland electric field, and magnetic-moment conservation. Starting from the Vlasov equation, we derive fully analytical expressions for the distribution functions, density, and parallel, perpendicular, and total temperature profiles. The combined conservation of energy and magnetic moment generates a loss-cone distribution, reducing the density relative to the unmagnetized case and producing temperature anisotropy. For a single-temperature boundary, the competition between gravity and magnetic-moment conservation produces a maximum in the parallel temperature, for which we derive and numerically validate analytical scaling laws. With stochastic heating, gravitational filtering enhances the contribution of hotter populations at coronal heights, while magnetic-moment conservation amplifies the velocity-space anisotropy. Our analytical solution provides a collisionless benchmark for future kinetic models incorporating more realistic magnetic-field geometries, Coulomb collisions, and turbulent particle scattering.

astro-ph.SR↗

arXiv2026

A solution to 2-copy distillability of Werner states

Entanglement distillation is a fundamental task in quantum information theory. In this work, we prove that Werner states in arbitrary dimension are $2$-copy distillable if and only if they are $1$-copy distillable. This answers the longstanding open question of the $2$-copy distillability of Werner states. This is an important step on determining whether every non-positive partial transpose (NPT) state is distillable, which remains one of the central open problems in the field of entanglement distillation.

quant-ph↗

arXiv2026

TILT: Model-Intrinsic Reward Alignment For Compositional Diffusion

Consider conditional generation $p(x \mid C=\{c_1, c_2, \dots c_k\})$ where $C$ is a prompt composed of multiple concepts $c_i$. Diffusion models often struggle with compositional prompts, producing samples in which some concepts dominate while others are missing or weakly represented. Prior work attributes these failures to mode collision, where single-concept modes of $p(x\mid c_i)$ overlap with modes of the joint $p(x \mid C)$. To seek out collision-free modes of $p(x \mid C)$, or "pure modes", corrector-based approaches have attempted to suppress collisions at intermediate diffusion times. However, local corrections are often heuristic and do not necessarily steer the generation to a "pure mode" in the final data space. Derived from a principled formulation, we present TILT (Test-time model-Intrinsic reward aLignment via Tilting), a training-free framework that poses eventual pure mode sampling as a reward for intermediate-time alignment. This reward offers valuable advantages: (1) it is intrinsic to the model, hence external reward models need not be trained by modality-specific datasets, (2) it yields a closed-form target under a variational approximation, which makes it realizable through standard diffusion sampling, and (3) it is interpretable, hence amenable to preference-based modifications. Project page: https://debottam-dutta7.github.io/tilt_web/

cs.AI↗

arXiv2026

Learning Lattice Parameters from Powder X-Ray Diffraction Data Using Invariants

We present a machine learning (ML) method to determine unit cell parameters from powder X-Ray diffraction (XRD) data using a novel invariant lattice representation. In ML, the data representation used can have a substantial impact on the prediction quality. Previous approaches have directly predicted lattice parameters ($a,b,c,α,β,γ$) from XRD inputs. However, these parameters depend strongly on the unit cell reduction or convention used. In this work, we construct an invariant representation of the reciprocal lattice that is independent of primitive cell convention, based on the bispectrum--a descriptor built from spherical harmonic projections of lattice points. The calculation of the lattice bispectrum is differentiable, and we demonstrate how to invert it using gradient-based optimization initialized from a nearest-neighbor lookup. We show that with a fixed ML model architecture, using the lattice bispectrum as the ML target rather than the unit cell parameters leads to more accurate lattice parameter predictions. For example, using the MP-20 dataset, the bispectrum reduces length mean absolute percentage error (MAPE) from 11.08% to 2.43% and angle MAPE from 12.15% to 2.45% compared to direct prediction with the same model architecture. We additionally benchmark our approach against pre-existing XRD to crystal structure models such as Crystalyze and assess its performance on experimental data. Beyond unit cell representation, we anticipate this invariant lattice representation could serve more broadly as a geometry-aware target for other crystallographic machine learning tasks such as structure generation.

physics.comp-ph↗