SEARCH · arXiv Science

Search arXiv Science

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

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 955 records · Page 53Linked to original sources

arXiv2026

Masked diffusion as random sequential adsorption: an exactly solvable model for parallel decoding

Masked diffusion models generate text by reversing a stochastic masking process. Unlike autoregressive LLMs, which generate text one token at a time, they can reveal many tokens in parallel, at the price of an error, since tokens revealed together are drawn independently. We show that this trade-off maps onto the statistical physics of driven many-particle systems. (i) For any data, the error of parallel generation is the entropy production of the generation process relative to the reversed masking process. (ii) For Markov data, generation is a random sequential adsorption process with labelled gaps, which we solve exactly for any number of steps. (iii) For many steps, the error becomes the dissipation of a slowly driven system: the unmasking schedule plays the role of the driving protocol, the number of steps that of its duration, and the error is set by a thermodynamic metric that we obtain in closed form from the mutual information between tokens. Optimal schedules are geodesics of this metric. The exact solution shows how correlations in the text shape both the error and the quality of the output. The longer the correlations, the more the error concentrates at the start of generation. Schedules that start with small steps then gain more, both in path error and in perplexity, and need fewer steps to reach their asymptotic behaviour.

cond-mat.stat-mech↗

arXiv2026

Anisotropic g-tensor and strong spin-orbit interaction of holes in hexagonal germanium nanowires

Hexagonal germanium combines direct-gap optical activity with the low-nuclear-spin environment of a group-IV material, making it a candidate platform for spin qubits and spin-photon interfaces. Using a ten-band $\mathbf{k}\cdot\mathbf{p}$ model, we study the response of confined hole states in hexagonal germanium nanowires to magnetic and electric fields. We show that, by symmetry, the top valence band of the bulk crystal has a vanishing $g$ factor in the plane perpendicular to the $c$ axis, in analogy with the in-plane $g$ factor of heavy holes in two-dimensional hole gases. In nanowires, confinement induces only small transverse $g$-tensor components, so the $g$ tensor remains strongly anisotropic. A transverse electric field strongly reduces the nanowire band gap, providing a gate-tunable optical signature and limiting the usable field range. The same field induces a Rashba spin-orbit interaction that is markedly stronger for fields perpendicular to the $c$ axis and depends nonlinearly on the field strength. In narrow wires at strong fields, the spin-orbit length reaches a few tens of nanometers, comparable to hole devices in cubic silicon. We identify the microscopic origin of the Rashba coupling as a second-order process mediated by a few intermediate subbands with distinct symmetry selection rules. Our results provide guidance for the design of electrically controlled hole-spin qubits in hexagonal germanium nanowires.

cond-mat.mes-hall↗

arXiv2026

Contrasting fate of holes introduced by Fe or In in kagome metal CoSn

CoSn is a kagome metal with well-defined flat bands 0.2 eV below the Fermi level EF. Moving them across EF would offer a rare opportunity to investigate their impact on electronic properties. Using angle-resolved photoemission spectroscopy (ARPES) and X-ray absorption at the Co K-edge, we examine two routes to introduce holes in this compound, through Co/Fe or Sn/In substitutions. Substituting In for Sn produces an almost rigid upward shift of the bands toward EF, as expected for hole doping, whereas substituting Fe for Co leads to an unexpected downward shift. Density functional theory calculations reproduce these opposite trends and trace the unusual behavior of Fe to its local magnetic moment. The moment keeps the hole on the Fe impurity and imposes an attractive electrostatic potential on neighboring Co atoms. Overall, both substitutions bring weakly dispersing states close to EF, the flat bands of the Co kagome lattice upon In substitution and partly localized states around magnetic Fe impurities upon Fe substitution.

cond-mat.str-el↗

arXiv2026

A Hypergeometric RM Surface: Integral Monodromy and Level-11 Covers

We study a genus-two hypergeometric family whose Jacobians have real multiplication by $\Q(\sqrt5)$. We identify the relevant primitive factor and integral rank-two monodromy; at one real embedding the projective image is $Δ(2,5,\infty)$, while the conjugate image is nondiscrete. At the primes above $11$, the degree-$12$ Borel covers are quadratic pullbacks of genus-one triangular modular curves. The same rank-two systems have finite-field hypergeometric Frobenius traces, yielding the local Euler factors and a quadratic transformation twisted by the character of $2$.

math.NT↗

arXiv2026

Learning Robotic World Models with Semantic and Visual Feedback

Robot world models predict action-conditioned future visual states to support robotic planning and decision making. However, visual reconstruction does not fully capture task progress, semantic evaluation alone provides limited constraints on fine-grained visual errors, and homogeneous predictions limit comparative feedback. We present RoboImagine, a framework for post-training autoregressive robotic world models through candidate-future comparisons with semantic and visual feedback. Diverse search promotes diversity among candidate trajectories under shared visual observations, action sequences, and task instructions. A frozen multimodal evaluator provides task-relevant semantic preferences through pairwise video comparisons in both presentation orders, complemented by pixel- and perceptual-level visual consistency feedback that constrains local structural and appearance errors. These signals drive policy optimization through group-relative advantages. Fixed initial context and branch-specific rolling histories support long-sequence generation with bounded context and consistent conditioning between candidate generation and policy optimization. Experiments with separately trained models on RT-1, BridgeV2, and RoboCasa cover real-world and simulated manipulation. On RT-1, under matched initialization, update counts, and evaluation protocols, RoboImagine improves PSNR by 1.26 dB and reduces LPIPS by 11.93% relative to continued supervised fine-tuning. Ablations show that diverse search improves prediction quality and that combined semantic and visual feedback yields higher external multimodal evaluation scores than visual feedback alone while maintaining comparable visual fidelity. Rolling histories further improve long-horizon prediction. These results support improving visual fidelity, task relevance, and long-horizon prediction through candidate comparison and complementary feedback.

cs.RO↗

arXiv2026

Long lived MeV scale bound state dark matter alleviates the ${}^7 Li$ tension

We investigate the impact of MeV scale unstable bound state dark matter on both thermal and non thermal Big Bang nucleosynthesis. Focusing on the weakly self interacting regime, we find that this framework can reproduce the theoretically predicted abundances through both short and long lived bound states. While long lived bound states can alleviate the cosmological ${}^7$Li problem, they simultaneously lead to discrepancies with other light element abundances. We show that this tension is resolved by incorporating non thermal nucleosynthesis, which act on these non standard initial light element abundances to alleviate the ${}^7 Li$ tension without violating observational constraints on the other elements. Remarkably, across the range of $\widetildeρ_B$ and $m_χ$ considered, long lived bound states with lifetimes $10^0\,{\rm s}\lesssimτ\lesssim10^2\,{\rm s}$ can simultaneously resolve the ${}^7$Li problem and a large parameter space remain consistent with the observed abundances of the other light elements. Our results demonstrate that non thermal effects are essential for obtaining a consistent nucleosynthesis history in unstable bound state dark matter scenarios.

hep-ph↗

arXiv2026

CANMentor: A Large-Small Model Collaboration Framework for CAN Bus Intrusion Detection

The Controller Area Network (CAN) bus lacks built-in security mechanisms, making real-time intrusion detection necessary. However, resource and latency constraints limit vehicle-side intrusion detection systems (IDSs) to lightweight models, whose detection performance and generalization can degrade under non-independent and identically distributed (non-IID) data. Meanwhile, uploading original CAN messages to the cloud for more powerful model inference risks exposing sensitive vehicle information. We propose CANMentor, a large-small model collaboration framework in which a cloud-side large language model (LLM) teacher supervises vehicle-side lightweight IDS models for CAN bus intrusion detection. Each vehicle first performs local inference, and a privacy-aware generator then produces proxies from the original CAN messages for cloud-side teacher inference to protect data privacy. The vehicle uses the teacher prediction only when the response arrives within a prescribed latency bound and otherwise retains the local prediction; all returned teacher logits are converted into soft targets for periodic knowledge distillation. On non-IID partitions of the ROAD dataset and the HCRL Car-Hacking dataset, knowledge distillation yields maximum relative Macro F1 improvements of 25.01% and 20.79%, respectively. Teacher predictions also improve detection accuracy for CAN messages near the local decision boundary by 38.71-42.00 percentage points.

cs.CR↗

arXiv2026

Towards Foundation Decision Models: Drex

Autoregressive approaches with large language models, such as long chain-of-thought reasoning, have become the default for new AI applications because they reach higher accuracy on a wide range of tasks. In production, however, they give inconsistent answers, require parsing of outputs, and add cost and latency. Decision models address this: given a question and a set of options supplied by the user, they return a probability distribution over the options without autoregressive decoding. We present Drex, a family of four open decision models with 8B (DLM), 9B (v1.5), 35B with 3B active (v2.0), and 27B (Max) parameters. Each request consists of a state (a document, record, or agent context, optionally with images) and questions about it, each asking for a choice among options, a yes/no judgment, or a score on an ordered scale. Training combines supervised fine-tuning with reinforcement learning from adjusted feedback (RLAF), plus vision tuning for Max. Ablations show that RLAF improves accuracy and that consistency losses improve robustness to option order. On Decision Index 0.4.1, Max sets a new state of the art among the decision models we compare (67.52 on the public suite vs. 60.70 for Jev 1.14.0). On Decision Bench v4, it is statistically tied with d3-turbo and GPT-6 Sol and trails Claude Opus 5.5 by 1.6 points, and it has the highest accuracy among decision models on DecideBench. Max answers a typical request in 183 ms, against about 2.0 s for GPT-6 Sol and 3.5 s for Claude Opus 5.5, and its outputs are stable under option reordering. All four models are open source: https://github.com/nace-ai/drex-decision-models

cs.CL↗

arXiv2026

Quantum Communication and Phase Sensing over a Dephasing Channel with Thermal Noise

Dephasing is a prominent source of decoherence in quantum communication, degrading phase coherence and transmission reliability. Therefore, knowledge of the physical channel parameters for reliable transmission is imperative, motivating the integration of quantum communication and sensing at the physical layer. Consequently, we study the transmission of coherent states over a von Mises dephasing channel affected by thermal noise. We consider a heterodyne receiver, for which, we first derive the symbol-error probability (SEP) and an accurate closed-form approximation, through which an irreducible asymptotic error floor due to the dephasing channel is identified. We note that the derived SEP is affected by the mismatch between the receiver's mean phase estimation and the true mean phase, thereby formalizing the interplay between sensing accuracy and communication reliability. In addition, we derive the Fisher information (FI) for the mean phase acquisition of the channel and provide closed-form lower and upper FI bounds. Moreover, the quantum FI (QFI) of the channel is studied, and lower and upper bounds on the QFI are given, which become asymptotically tight as the photon number increases. Despite the initial complexity of the FI and the QFI, the bounds given are relatively simple and effective. In addition, by comparing the FI with the QFI, we show that heterodyne detection approaches the quantum sensing limit as the number of photons increase, while an unavoidable sensing floor exists due to dephasing. Then, using the SEP and the FI analysis, we discuss the conditions under which communication and phase estimation requirements are jointly feasible. Finally, numerical results validate the derived SEP, estimation bounds, and the sensing-communication feasibility tradeoff.

cs.IT↗

arXiv2026

Three-critical hypergraphs of maximum local edge connectivity two

For a hypergraph $G$, let $χ(G)$ denote its chromatic number and $λ(G)$ the maximum, over pairs of distinct vertices, of the minimum size of an edge cut separating the pair. A hypergraph is $3$-critical if its chromatic number is three and every proper subhypergraph is two-colorable. In this paper, we give a recursive characterization of the $3$-critical hypergraphs with $λ(G)\le2$. We prove that these are precisely the hypergraphs generated from ordinary odd cycles and an explicit family of labeled-tree hypergraphs by admissible Toft splits. This characterization includes a necessary and sufficient condition for a prescribed split to preserve $λ\le2$. The condition involves only cuts in the first factor and the two traces on the split edge. More generally, we establish an exact cut formula for splits at degree-$k$ vertices of $(k+1)$-critical hypergraphs, for every $k\ge2$. We further prove that the smallest member of this critical class outside the family generated by hyperwheels and both forms of the Hajós join has eight vertices and is unique up to isomorphism. Together, these results resolve two issues left open by Schweser, Stiebitz, and Toft in \emph{Journal of Combinatorics} (2022): the characterization of hypergraphs with $λ(G)=2$ and $χ(G)=3$, and whether hyperwheels and Hajós joins suffice to generate all $3$-critical hypergraphs in this connectivity range.

math.CO↗

arXiv2026

Liouville systems I: Blow-up analysis and orthogonal decomposition

We study blow-up solutions of generalized Liouville systems on a closed surface at and near critical parameter hypersurfaces. Under a Morse assumption on the reduced functional, we establish the following results. First, we introduce amplitude variables in the approximate manifold and prove that every sequence of blow-up solutions at a fixed critical parameter, or along an admissible sequence from the lower side, admits a unique minimizing decomposition up to permutation. The remainder converges to zero and is orthogonal to the center, scale, and amplitude directions. The concentration scales are comparable, the limiting configuration satisfies the scale balance conditions, and the amplitudes vanish after multiplication by the logarithmic scale. Second, we determine the leading parameter defect in both critical decay regimes and prove that, along each fixed admissible path, all sufficiently large solutions lie in finitely many disjoint bubbling neighborhoods. Finally, assuming the leading coefficient is nonzero at every balanced critical configuration, the solution set at the fixed critical parameter is compact. These results provide the analytic foundation for studying the resonant degree via critical points at infinity.

math.AP↗

arXiv2026

MSR-CP: A Modified Segmented Regression Approach to Change Point Estimation in Rainfall Patterns in West Bengal

This study presents a methodology for detecting change point in precipitation series using a smooth version of segmented regression model. We examine several existing change point detection methods and identify their limitations in capturing gradual changes commonly observed in climate data. To address these limitations, we discuss an extension of the traditional segmented regression model enabling the detection of gradual shifts in trend. The developed methodology is further extended to accommodate dependent error structures, which are common in time series data. Extensive simulation studies are conducted to evaluate the performance and computational efficiency of the proposed approach relative to existing methods. The methodology is subsequently applied to 60 years of monsoon rainfall data from two districts of West Bengal, India. This approach offers improved insights into evolving rainfall patterns and serves as a useful tool for climate data analysis and long-term forecasting.

stat.ME↗

arXiv2026

CORE: Coarse-to-Fine Search with Context Management for Deep Research

Deep research agents answer complex questions through repeated search. They rely on a harness that runs their tool calls and returns evidence to their context. This creates a trade-off: Search works best over long passages, which give enough context for reliable semantic matching. But the returned evidence should be short, because every observation stays in the agent's context for the rest of the trajectory. We introduce CORE, a training-free harness that combines coarse-to-fine search with context management. CORE retrieves passages with a dense retriever. It then reuses the token-level similarities from a late-interaction reranker to return only the most relevant snippet from each passage. Context management lets the agent expand the context around sources it has already retrieved and suppresses repeated snippets. In controlled closed-search comparisons with the Tongyi DeepResearch and OpenResearcher harnesses, CORE achieves state-of-the-art answer accuracy and evidence recall across three agent models and three benchmarks (BrowseComp-Plus, BrowseComp-Wikipedia and BrowseComp-Plus-CM), on corpora containing up to 553M documents.

cs.CL↗

arXiv2026

Extending graph metrics to hyperspaces of vertex sets: optimization and reinforcement learning applications

An undirected connected graph can be interpreted as a metric space by considering the path distance between its vertices. Extending this perspective, we explore how walks in the graph, viewed as sequences of subsets of vertices $V$, fit into a more general framework where the path distance emerges as a special case of broader metrics that define hypertopologies on the space of subsets of $V$. In this paper, we propose a general framework to address this issue, motivated by the extension of recent reinforcement learning algorithms on graphs through Lipschitz function extensions. Our approach demonstrates how models for dynamical systems and related machine learning techniques can be adapted to (quasi-pseudo-metric) hyperspaces, offering a novel perspective for the analysis and modeling of dynamical systems. Finally, we illustrate the practical relevance of our theoretical contributions through a concrete graph optimization example, highlighting the utility and effectiveness of our proposed approach.

math.GN↗

arXiv2026

Continuous variable payload transmitter for the SPOQC mission

The continuous variable (CV) quantum key distribution (QKD) scheme enables two parties to share a secret key separated by a certain distance. Here we propose, design, build, test and show results of a CV payload transmitter, intended to demonstrate downlink based QKD from a CubeSat in low earth orbit. This is part of the SPOQC mission by Quantum Communications Hub, UK which was successfully launched on March 2026.

quant-ph↗

arXiv2026

On the aerodynamics of an accelerating and decelerating propeller in hover at low Reynolds numbers

Small unmanned aerial vehicles commonly regulate propeller thrust through rapid changes in rotational speed, yet it remains unclear how the inflow, the wake, and the near-surface velocity respond to these changes at low Reynolds numbers. A commercial two-bladed propeller was tested in hover with custom blade holders at three additional pitch offsets relative to the original blade geometry, $Θ=\,$0, 5, and $10^\circ$, and accelerated and decelerated between 3000 and 5000 RPM with open-loop step responses. The resulting nondimensional acceleration and deceleration numbers, below 0.011 and 0.007, suggest negligible acceleration effects. Phase-locked particle image velocimetry (PIV) at 40, 60, and $80\,\%$ rotor radius and in the meridian plane was combined with load measurements. The thrust coefficient responds quasi-steadily at $Θ= 10^\circ$, but shows hysteresis at the higher pitch offsets, with lower values during acceleration and higher values during deceleration than under steady conditions. This trend opposes expectations from dynamic inflow theory. The effect is largest at $Θ= 5^\circ$, with deviations from the steady values of up to 8\,% during acceleration and $5\,\%$ during deceleration, compared with $4\,\%$ and $3\,\%$ at $Θ= 10^\circ$. PIV reveals that the rotor-disk inflow, blade deformation, and sectional circulation remain largely quasi-steady, whereas velocity fluctuations, blade-wake vorticity, and tip-vortex circulation vary strongly, with the largest deviations at low rotational speeds rather than at peak acceleration or deceleration.

physics.flu-dyn↗

arXiv2026

Local and 2-Local Derivations of Finite-Dimensional Zinbiel Algebras

We study local and $2$-local derivations of finite-dimensional Zinbiel algebras. Over any field of characteristic zero, every Zinbiel algebra $A$ with $A^2 \neq 0$ and $\dim A \leq 4$ admits a proper local derivation. The proof uses a filtration argument. We exhibit two five-dimensional Zinbiel algebras with nonzero multiplication whose derivations are all central. Every local derivation on either algebra is therefore a derivation, so the minimum dimension for local rigidity with nonzero multiplication is five. Annihilator-valued interpolation gives a general construction of proper local derivations. When $A^3 = 0$, the multiplication tensor describes all derivations and local derivations. For a finite-dimensional nilpotent algebra $A$ over any field, we show that every $2$-local derivation is a derivation if and only if $\dim(A/A^2) \leq 1$. The proof combines the annihilator interpolation of Sh. Ayupov, K. Kudaybergenov and B. Omirov with a one-generator argument. When all derivations are central, every $2$-local derivation is obtained by composing a homogeneous map $A/A^2 \to \text{Ann}(A)$ with the quotient map, and every such composition is a $2$-local derivation. The five-dimensional examples extend to locally rigid Zinbiel algebras with nonlinear $2$-local derivations in every dimension at least five. We also show that derivations given by left multiplication form a Lie ideal of nilpotency class at most two and give sufficient conditions for the existence of left outer derivations. This ideal is zero in our locally rigid examples.

math.RA↗

arXiv2026

Quark-mass dependence of two-nucleon scattering

Understanding how the properties of the visible Universe depend on the parameters of the Standard Model of particle physics requires \emph{ab initio} theoretical computations. To that end, we investigate the quark-mass dependence of low-energy nucleon-nucleon scattering in the deuteron and dineutron channels. We perform lattice-QCD calculations at pion masses $M_π\approx 200$ and $280$ MeV and compare their results to phenomenology at the physical point and to a previous calculation at $M_π\approx 710$ MeV. Scattering amplitudes, phase shifts, and threshold parameters are determined from finite-volume spectra using quantization conditions that explicitly incorporate long-distance effects from one-pion exchange. We analyze the pion-mass dependence of scattering lengths and effective ranges using chiral effective field theory with Weinberg and Kaplan-Savage-Wise power countings. From this, we estimate the pion-mass derivatives of the inverse scattering lengths and binding momentum, at the physical point. Both power countings indicate that the deuteron unbinds near $M_π\approx 220$ MeV.

hep-lat↗