Papers

The under-the-barrier recollision pathway during the tunneling ionization of an atom, despite its minuscule amplitude, leaves its signature on the observable photoelectron momentum distribution due to a phase shift relative to the direct ionization path, that can be interpreted as a tunneling time delay. We investigate how this time delay is modified in the nonadiabatic regime, where the tunneling barrier evolves dynamically during the ionization process. Our analysis is based on a simple model of a one-dimensional short-range potential driven by a strong laser pulse which remains amenable to a systematic analytical treatment with the strong field approximation (SFA). We show a significant contribution of the high-order SFA corrections, which describes the recollision channel with recombination and further ionization, and is additional to the rescattering. Whereas the recombination channel dominates over the rescattering for the tunneling time delay in the quasistatic regime at relatively large laser fields, their contributions are reversed in weak fields.

We develop a deformation-theoretic framework for algebraic superstacks and apply it to the construction of perfect obstruction theories on moduli spaces of stable supermaps. In order to achieve this we employ the tools and concepts from derived algebraic geometry, extending them to the supergeometric setting. Our main technical contributions are (i) the notions of derived superstacks, (ii) their cotangent supercomplexes and (iii) perfect obstruction theories for algebraic superstacks. We prove that a quasi-smooth derived enhancement of an algebraic superstack always yields a perfect obstruction theory: this generalizes a well-known principle from classical deformation theory. We then turn to the case of the Deligne–Mumford superstack $\boldsymbol{\mathcal{M}}^{\mathsf{st-}\mathrm{SUSY}}_{g,\mathfrak{n}}(\boldsymbol{Y})$ of stable supermaps of fixed genus, homology class and numbers of Neveu–Schwarz and Ramond–Ramond punctures, with target a smooth projective superscheme $\boldsymbol{Y}$. We exhibit a natural derived enhancement $\mathbb{R} \boldsymbol{\mathcal{M}}^{\mathsf{st-}\mathrm{SUSY}}_{g,\mathfrak{n}}(\boldsymbol{Y})$ and compute its cotangent supercomplex; we prove that the induced obstruction theory on $\boldsymbol{\mathcal{M}}^{\mathsf{st-}\mathrm{SUSY}}_{g,\mathfrak{n}}(\boldsymbol{Y})$ is perfect. Finally, we compute its virtual dimension by super Grothendieck–Riemann–Roch. The resulting formula gives, in particular, a deformation-theoretic proof of the virtual dimension formula previously conjectured for stable supermaps and specializes to the usual virtual dimension of the classical moduli stack of stable maps.

The increasing use of lattice-like architectures in load-bearing members, facilitated by advances in additive manufacturing, calls for efficient multiscale approaches to their structural analysis. For slender lattice-like structures, beam models provide a computationally efficient macroscopic representation while retaining the ability to account for the underlying lattice architecture through an homogenization approach. Although finite element squared (FE2) approaches combining beam models at the macroscopic scale with continuum models at the microscopic scale have been established, an FE2 framework combining a beam-based macroscopic model with a truss-based microscopic model of a lattice structure has not yet been reported. To address this gap, an FE2 framework is developed for three-dimensional shear-deformable beams with periodically repeated lattice-like microstructures and periodic displacement boundary conditions. The proposed FE2 framework introduces constraint equations that (a) prevent rigid-body translations and rotations of the representative volume element (RVE) without imposing additional kinematic restrictions and (b) ensure that the homogenized stress resultants and material matrix are independent of the selected RVE length. To mitigate artificial boundary effects arising at the boundaries of the lattice-like RVE, local correction factors are introduced. The framework is assessed through numerical studies of two- and three-dimensional lattice-like structures with different geometries, considering both geometric and material linearity and nonlinearity. The results demonstrate the independence of the homogenized stress resultants and material matrix from the selected RVE length and show good agreement with the corresponding reference models in all investigated cases.

A light neutrinophilic scalar $\phi$, which couples predominantly to neutrinos, can be probed through leptonic meson decays for scalar masses $10~\mathrm{MeV}\lesssim m_\phi\lesssim\mathcal{O}(100)~\mathrm{MeV}$. We study neutrinophilic-scalar emission in $D$-meson decays at BESIII and the future Super Tau-Charm Facility (STCF), focusing on $D_{(s)}^+\to e^+\bar{\nu}_\alpha\phi$. Accounting for the $m_\phi$-dependent missing-mass squared ($M_{\rm miss}^2$) distribution within the experimentally selected region, we find the resulting $D^+$ constraint to be substantially weaker than previous estimates based on the integrated three-body decay rate. Extending the $M_{\rm miss}^2$ coverage provides sensitivity to heavier scalars. With the full kinematic range of $M_{\rm miss}^2$ and an integrated luminosity of $1~\mathrm{ab}^{-1}$, the projected STCF sensitivities surpass the existing kaon bound for $m_\phi\gtrsim 380~\mathrm{MeV}$ in the $D^+$ decay and $m_\phi\gtrsim 310~\mathrm{MeV}$ in the $D_s^+$ decay. We further discuss direct searches for $D_{(s)}^+\to\mu^+\bar{\nu}_\alpha\phi$ and their potential for probing the neutrinophilic scalar. A complete analytic expression for the three-body decay width is also derived with full dependence on the scalar and charged-lepton masses, including its behavior in the $m_\phi\to 0$ and $m_\ell\to 0$ limits.

The Erdős–Gallai theorem determines the maximum number of edges in a graph with bounded matching number; its clique-counting extension replaces edges by $s$-cliques, and a spectral analogue in terms of the $s$-clique tensor has recently been established. We study the corresponding $\mathcal A_\alpha$-tensor of the $s$-uniform clique hypergraph. For $0\leq\alpha\leq1$ and $3\le s\le2t-1$, we determine the maximum $\alpha$-$s$-clique spectral radius among $n$-vertex graphs containing no matching of $t$ edges: when $3\le s\le t$ and $n$ is sufficiently large, the maximum is attained by the join of a clique of order $t-1$ and an independent set, and when $t<s\le 2t-1$ and $n\ge2t-1$, it is attained by a clique of order $2t-1$ together with isolated vertices. At $\alpha=0$, these statements recover the known result for the $s$-clique spectral radius; at $\alpha=1$, they yield the corresponding statement for the maximum $s$-clique degree. For $s=t\ge3$, we obtain the maximum for every $n\ge2t-1$; at $\alpha=0$, this removes the requirement that $n$ be sufficiently large from the known result.

Automatic Speech Recognition (ASR) systems often show uneven performance across demographic groups, and errors can be especially difficult to address for speakers belonging to multiple demographic groups. This work studies demographic-aware model merging for fair Speech-LLM-based ASR. Starting from a SLAM-ASR-based model, we fine-tune only the connector on demographic-specific subsets and merge the resulting subgroup-adapted connectors into a global model. We then identify critical cross-axis demographic pairs using subgroup WER and task-vector conflict, and apply intersection-specific correction vectors to the global merged model. Experiments on Fair-Speech show that global demographic merging improves overall WER over the base model, while intersection correction provides additional gains for several merging strategies. In particular, TIES with WER-based correction achieves the best overall WER, reducing it from 7.38% to 5.13%. Subgroup and disparity analyses further show that the proposed approach improves performance across demographic axes, while highlighting that lower average WER does not always imply reduced subgroup disparity.

Early stress detection in crops is a necessity today to improve efficiency and reduce waste of time, money, and effort. However, most modern techniques, such as hyperspectral imaging and AI-based systems, are too costly and complex for medium and small-scale farmers to implement. This paper showcases CropSentry, a low-cost, ground-based multi-robot system that uses multimodal leaf sensing to continuously monitor crop health by tracking stress levels. The system comprises two autonomous bots that continuously detect leaf color and environmental data row by row. The observations are spatially mapped and sent over to the master bot, which uses color-coded row segments to generate a real-time web-based dashboard displaying crop health. After 63 observations were collected during the experiments, the results showed an overall crop health classification accuracy of 84.12%, with 82.60% for healthy plants, 88% for nutrient-deficient plants, and 80% for diseased plants. Also, 100% wireless communication success rate across 10 slave observations was achieved. Close-range leaf inspection across multiple bots can detect early stress in crops while remaining affordable, accessible, and scalable. It provides farmers with timely information to improve resource utilization and crop management.

Split conformal prediction sets for competing-risks labels at a fixed horizon require calibration labels that right censoring can leave unobserved. Complete-case calibration guarantees coverage for the label-complete subpopulation, but its population coverage can deviate in either direction, even at the true class probabilities. We study how selection changes the score distribution near the population quantile. At the true score in our main simulation family, with independent draws, complete-case calibration covers 0.8723 at a nominal 0.900 when 22% of subjects are event-free at the horizon. In 4 of 48 further designs using per-draw normalisation, estimating cause incidences as one minus the exponential of the negative of the cause-specific Nelson-Aalen cumulative hazards, with the event-free probability as the clipped and renormalised remainder, puts complete-case coverage at least four standard errors below nominal. In these designs, weights from a correctly specified censoring model keep coverage near nominal. Extending the argument of Yi et al. (2025) to a cause label, we establish a finite-sample coverage lower bound with an explicit penalty for censoring-model error. Misspecifying the censoring model can also lead to under-coverage.

Emergency messaging such as extreme-weather reports and earthquake instructions can involve high stakes, to the extent that translation errors can lead to tragic consequences. The use of machine translation or generative artificial intelligence might therefore not be recommended. On the other hand, time savings in the initial translation can allow greater investments of resources in revision and authorization processes, as well as a wider range of target languages. An experiment with generative AI translations of an earthquake instruction text from English into Chinese and Spanish shows that use of discourse-specific prompts can considerably improve understandability and actionability, although the translations may still not be trusted by translators. Human revision is still required, not only to detect errors but also because of the ethical need for someone to take responsibility for any errors or delays in such messaging.

We study the $d$-dimensional generalization of Models II ($d=2$) and V ($d=3$) for liquid crystals introduced by Heilmann–Lieb (1979), which consists of dimers on the hypercubic lattice at chemical potential $\mu$, interacting via a hard-core repulsion and a lateral attraction between parallel dimers with coupling constant $b>0$. For $d\in\{2,3\}$, these models were conjectured to exhibit liquid-crystalline order at low temperatures provided that $\mu+2(d-1)b>0$. We establish nematic order for all $d\ge 2$ under the additional condition $(4d-6)b>\mu$, which corresponds to the physical regime in which misaligned dimers are rare on scales comparable to the correlation length of the one-dimensional reference system. While the treatment of orientational order here follows the approach of our recent work on Model I, the proof of the absence of translational order relies on a custom implementation of cluster swapping in the sense of Sheffield (2005). Consequently, we deduce uniqueness of Gibbs measure and anisotropic exponential decay of correlations within each oriented phase.

Mandal and Mehatari (Comp.\ Appl.\ Math., 2025) introduced the class $\mathcal{C}$ of cographs generated by a finite creation sequence $(\alpha_1,\dots,\alpha_m)$ of natural numbers, and derived the inertia, an extended eigenvalue-free interval, and the exact characteristic polynomial for the adjacency matrix of such graphs. In this note we develop the parallel theory for the distance Laplacian matrix $D^L(G)$ and distance signless Laplacian matrix $D^Q(G)$. It is shown that $(0,\alpha_{\min}) \cup (n-\alpha_{\min}, n)$, $(0,n) \ \cup\ (n,\,n+\alpha_{\min}) \ \cup\ (2n-\alpha_{\min},\,2n)$, $\big(2\Tr_{\min},\ \mu_{\max}(D^Q(G))\big)$ are the eigenvalue-free intervals of the Laplacian, distance Laplacian, and distance signless Laplacian matrices respectively for the said class of graphs.

We discuss free-energy estimation for complex systems using simulated tempering with Wang–Landau-type feedback. We show that introducing nonreversible dynamics in temperature space and multiple temperature updates per step significantly accelerates the convergence of the free-energy estimate. We then apply the algorithm to the three-dimensional Edwards-Anderson spin-glass model. Our results show that the algorithm can precisely estimate the free energy up to a large system size $L \simeq 64$, and that subsequent equilibrium simulated tempering simulations with the estimated free energy can rapidly sample from the Boltzmann distributions over a wide temperature range that covers the spin-glass transition temperature. The convergence time of our algorithm grows much more slowly with system size than that of another known algorithm for on-the-fly free-energy estimation.

In a Friedmann universe with a time-varying cosmological term, the production of massive particle–antiparticle pairs and their exchange of energy with the spacetime vacuum are described by a cosmic rate equation, $\dot\rho_M+3H\rho_M=\Gamma_M(\rho_M^H-\rho_M)-\Gamma_M^{\rm de}\rho_M$. Together with the Friedmann equations, an energy-balance equation for the time-varying cosmological term, and a reheating equation for the radiation, it forms a closed set of four independent background equations. We derive this set from a complex scalar field with a global U(1) symmetry and a non-minimal coupling to curvature. In the adiabatic case, the occupation of non-relativistic modes gives $\rho_M^H=2\chi m^2H^2$. A kinetic equation, based on the pairwise coupling $\xi R|\Phi|^2$ to fluctuations of the spacetime vacuum, gives the relaxation rate $\Gamma_M$, and a Yukawa coupling gives the decay rate $\Gamma_M^{\rm de}$. In the fast-oscillating case, the Hubble function oscillates at twice the pair mass and locks in phase with the pairs, so that it exchanges no net energy with them. The energy of the created pairs is supplied instead by the decrease of the cosmological term in the time evolution, $-\dot\rho_\Lambda=[(3-2\epsilon) H+\Gamma_M^{\rm de}]\rho_M^H$. Averaging over the fast time gives the cosmic rate equation in the detailed-balance regime $\Gamma_M\gg H$. In this regime the closed set has an exact power-law solution with $\epsilon=-\dot H/H^2=\chi m^2/m_{\rm pl}^2$. We discuss the application of the closed set to inflation, reheating and standard cosmology.

We argue that some of the puzzling features associated with the normalization of gravitational wave functions can be substantially alleviated by restricting the path integral to a sum over connected, allowable geometries. This prescription gives sensible results when applied to classical transitions and it provides non-trivial no-boundary probabilities, in line with old expectations. We highlight the additional impact of both boundary conditions and integration contours on these issues.

Deep generative models often produce structured outputs composed of interacting components. Modelling these outputs with a single model requires learning both the component distributions and their interactions. We pursue a modular alternative: reuse independently trained component generators and learn only how to coordinate them to produce coherent structured outputs. Our framework, Coordinated Multi-Agent Diffusion Steering (CMDS), treats frozen pretrained diffusion models as reusable generative primitives and coordinates their reverse processes through a learned control. We formulate coordination as a stochastic optimal control problem, balancing an assembly-level reward that specifies the desired properties of the combined output against deviations from the pretrained dynamics. The learned control amortises this optimisation, allowing reuse across new task instances. Experiments show that CMDS can recover a known target distribution, satisfy different spatial constraints with the same trained control, and recover individual sources from degraded mixtures. Across multi-agent maze navigation, articulated robot planning, and text-conditioned human motion, CMDS turns frozen models into coordinated multi-agent generators.

We prove asymptotic completeness of the wave operators in time-dependent scattering theory which intertwine the ordinary or fractional Laplacian and the corresponding Hardy operator, i.e., the fractional Laplacian with an added subcritical or critical Hardy potential, whenever the Hardy potential is short-range. Moreover, we prove asymptotic completeness of the wave operators intertwining Hardy operators with and without external short-range perturbations. We also show the nonexistence of the considered wave operators when the Hardy or external potentials are long-range. We arrive at these results using three different approaches. The first approach uses an abstract semigroup framework for asymptotic completeness of wave operators for Fourier multipliers with short-range potentials. The second approach uses Kato smoothness theory. In the third approach, we develop and apply new propagation estimates for Hardy operators with and without perturbations.

Traditional approaches for automated bug detection in video games, such as manual testing, can be beneficial for the improvement of quality assurance, but they can be expensive and time-consuming. The scarce number of tools available to detect multiple perceptual bugs in the same video frame introduces detection challenges for automated bug detection tools in real-world scenarios. We propose a deep learning model for multi-label perceptual bug detection and compare it against video classification models such as Inflated 3D ConvNet and 3D ResNet. Our proposed model, ResNet-BiLSTM, achieved an F1 score of 85.78% on the benchmark dataset. Our results demonstrated that temporal dependency modelling is beneficial for accurate video-based bug detection. We believe this work with multi-label perceptual bug detection on gameplay videos will help save resources spent on manual testing workloads in video games. Furthermore, we introduce a new dataset with multi-label perceptual bugs in this work. The dataset contains 77,969 video clips across different genres of games with approximately 1.2 million frames, containing combinations from 5 classes of bugs in the same video frame.

CNet is a C++/CUDA framework for deep complex-valued neural networks (CVNNs), trained by gradient descent with explicit Wirtinger (CR-calculus) derivatives. Complex models are underexplored yet natural where data is intrinsically complex – RF/IQ, MRI k-space, radar/SAR, audio spectra – and phase carries information real networks discard. CNet is physics-native: a network is a cascade of complex (often unitary) operations on an amplitude vector, and classification is a Born-rule measurement p_k = |z_k|^2 / ||z||^2 rather than a softmax over real logits. Every layer ships a CPU reference and a finite-difference-verified CUDA kernel; the static, batch-cloned computation graph captures cleanly into CUDA Graphs, outperforming eager and compiled torch.complex64 on an H100. We present three studies. (1) A fully complex causal language model, built on a new causal (triangular-masked) Fourier mixer with an O(N log N) Bluestein evaluation of the forward pass and its Wirtinger adjoint, matches a parameter-matched real causal FNet on character-level tiny-shakespeare in under half the steps. We introduce Born-rule attention, a content-based causal mixer whose weights are the quantum-measurement probabilities |<Q_k, K_j>|^2 – to our knowledge the first classical, complex-valued attention to replace softmax with the squared-magnitude Born score; with rotary embeddings and per-token complex normalization it reaches 1.52 nats/char, matching softmax self-attention, and transfers unchanged to byte-level (1.31 nats/char) and sub-word WikiText-103. (2) On radio-modulation classification (RML2016.10a/RML22) we trace the pooled-power accuracy ceiling to discarded temporal structure and recover it with a learnable cyclostationary conjugate-lag operator. (3) A bottleneck analysis of the Fourier phase problem in coherent-diffraction imaging isolates where complex-valued networks still need new operators.

Local-basis Shannon entropies of critical wave functions contain universal subleading information, but the Shannon point of the critical transverse-field Ising chain is singular in the conventional Rényi approach. We treat it directly at $n=1$ by representing the complete computational-basis Born distribution as the odd-degree boundary of independent long-range Bernoulli edges. The Shannon chain rule separates the entropy into explicit independent-edge and conditional cycle-space terms. For the periodic chain this isolates the analytic edge anomaly and organizes the remaining constant by linked vertex support. Conditional on the stated finite-part matching and forest assumptions, every nonvanishing linked coefficient is represented by an explicit finite-dimensional integral, giving an all-orders analytic hierarchy; the four-vertex term is evaluated in closed form, and the first few linked sectors already nearly saturate the established periodic constant, which is also reconstructed independently from the exact finite-size distribution. For an interval, the exterior reduces exactly to one ghost vertex and the explicit squared logarithms cancel, leaving the analytic edge contribution $\gamma_E=\log2/4-1/16$. The conditional cycle term has an exact all-support decomposition into physical-line and ghost-linked sectors; conditional on closure of the boundary forest subtraction, these sectors define an all-orders hierarchy of renormalized line and monomer–dimer boundary periods for the remaining logarithmic coefficient. The first complete augmented linked coefficient is evaluated in closed form, with an exact cancellation between its line and ghost boundary-layer anomalies; an independent reconstruction from exact finite-size probabilities recovers the established coefficient $\gamma_1=0.060020(3)$.

Learning-based semantic communication is vulnerable to adversarial perturbations introduced before semantic encoding or over wireless channels. This paper proposes TwinViT-DeepJSCC, a preventive-corrective semantic image transceiver operating under a fixed channel-use budget. Two Vision Transformer (ViT)-based deep joint source-channel coding (DeepJSCC) branches learn complementary latent representations protected by sensitivity-aware masking. At the receiver, confidence-aware fusion, blind corruption-severity estimation, and signal-to-noise ratio (SNR)-severity-conditioned denoising diffusion implicit model (DDIM) purification mitigate residual corruption without requiring attack metadata. Experiments on the Canadian Institute for Advanced Research 100-class (CIFAR-100) dataset consider fast gradient sign method (FGSM), projected gradient descent (PGD), natural evolution strategies (NES), and Carlini-Wagner (CW) source-domain attacks, as well as random jamming and channel-aware adversarial waveforms over additive white Gaussian noise (AWGN) and block-flat Rayleigh fading. Under matched channel-use and attack budgets, TwinViT-DeepJSCC achieves maximum peak signal-to-noise ratio (PSNR) gains of approximately 9.5 dB under 20-step PGD and 10.8 dB under channel-aware waveform attacks over block-flat Rayleigh fading. Under PGD, it also improves Top-1 accuracy by up to approximately 38 percentage points over the undefended baseline and 13 percentage points over the strongest competing defense. Ablation results confirm the complementary contributions of the proposed transmitter- and receiver-side mechanisms.

We study the Gieseker moduli space $\mathcal{M}_{X_0}$ of stable rank-$2$, degree-$1$ vector bundles on an irreducible curve $X_0$ with one node. Its normalization is a two-step blow-up of the moduli space of generalized parabolic bundles on the normalization $\tilde X_0$, and $\mathcal{M}_{X_0}$ is recovered by gluing along a natural involution on the boundary. From this we compute $\operatorname{Pic}(\mathcal{M}_{X_0})$ and, via a Harder–Narasimhan stratification, the virtual Poincaré polynomial of $\mathcal{M}_{X_0}$. Since the normalization morphism is small, the intersection cohomology of $\mathcal{M}_{X_0}$ equals the ordinary cohomology of the smooth normalization, which gives a closed formula in terms of the arithmetic genus.

Metric theories of gravity admit a massless breathing scalar $b$ besides the tensor modes $+$ and $\times$. Many scalar–tensor models predict a much smaller breathing amplitude; gravitational quantum field theory (GQFT) instead allows $h_b$ comparable to the tensor amplitudes. We evaluate the LIGO–Virgo–KAGRA (LVK) antenna patterns with PyCBC, including Michelson suppression of $b$. Here $F_b$ is the breathing antenna pattern and $\mathcal{S}_b=\sum_d F_b^{d2}$ is the network breathing power. At the H1 zenith at the GW150914 epoch, $F_b\sim 5\times 10^{-6}$; for $72$ Gravitational-wave Transient Catalog (GWTC) events with packaged strain the median inclination gain in $\mathcal{S}_b$ is $\sim 34\times$. Ranking multi-detector GWTC events (network $\mathrm{SNR}\ge 8$, $\ge 2$ sites) by $\mathrm{SNR}\sqrt{\mathcal{S}_b^{*}}$ identifies loud O4 triggers and GW170817 as geometric targets. Parameter-estimation (PE) weighted $\mathcal{S}_b$ on twelve events shows that the true sky can suppress this ranking (GW170817) and that GW200129_065458, GW150914, and GW190814 have the highest PE-weighted $\mathcal{S}_b$ for $\xi$ in $h_b=\xi h_+$ (GR: $\xi\equiv 0$). An HLV design-noise injection grid gives $\ln B(\mathrm{GR}+b/\mathrm{GR})\approx 13$ ($4$) for $\xi=1$ ($0.5$) on an inclined sky ($\mathcal{S}_b\sim 0.18$); near zenith ($\mathcal{S}_b\sim 0.016$) the $\xi$ posterior remains unconstrained. The GWTC limits are semi-joint: masses and spins fixed at PE medians, while $d_L$, $\theta_{JN}$, sky, and $\xi$ vary within PE-informed priors. Eleven of twelve events are analyzed (GW170817 omitted: PE sky on the breathing null). None favors $\mathrm{GR}+b$; $|\xi|_{90\%}\approx 0.15$, $0.18$, and $0.23$ for GW150914, GW200129_065458, and GW200311_115853 disfavor $|\xi|\sim\mathcal{O}(1)$, while weaker bounds (including GW190814) still allow that scale.

We study primal and dual error bounds for convex composite optimization problems and their Fenchel duals. For the primal problem, we use a proximal-gradient residual, while for the generally nonsmooth dual problem we use a subdifferential residual. We introduce a reduced primal–dual KKT mapping whose zero set coincides with the Cartesian product of the primal and dual solution sets under the standing solvability and strong-duality assumptions. We establish implications in both directions between metric subregularity of the reduced KKT mapping and the primal and dual local error bounds. If the composite matrix has full row rank and the smooth gradient is locally Lipschitz continuous, KKT metric subregularity is equivalent to the primal local error bound. If the smooth term is strongly convex, it is equivalent to the dual local error bound. Under full row rank and strong convexity, KKT metric subregularity is equivalent to the simultaneous validity of both error bounds. We also establish relations between the Luo–Tseng and local error bounds. The framework is then applied to quadratic–polyhedral models, strongly convex smooth models with general convex regularizers, and regularized least-squares models beyond polyhedrality and strong convexity, providing a unified treatment of primal, dual, and KKT error-bound properties.

Long conversational agents have become essential in our daily lives. They must remember what was said long back in order to help us efficiently complete a task without needing the user to repeat instructions and context repeatedly. However, the main issue is that instructions and context change over time and so the agents must be able to adapt accordingly. A useful memory system should preserve both current and historical states, distinguish stale information from active knowledge, retrieve evidence appropriate to the query and avoid repeatedly invoking a large language model to rewrite prior interactions. We introduce MINDSET, a memory controller that stores a conversation as immutable episodes and organizes them into versioned schemas through minimum-energy state transitions. Each incoming episode may reinforce, supersede, split or create a schema. The transition decision balances representation distortion, contradiction, historical damage, fragmentation and internal inconsistency, while hysteresis prevents isolated contradictions from prematurely rewriting stable memory. We evaluate MINDSET against 5 memory systems on a reproducible sample of 850 questions (700 LoCoMo + 150 MemoryAgentBench). MINDSET obtains the highest observed LoCoMo answer F1 while significantly improving retrieval ranking (Recall@8, MRR and nDCG@8) over the second best method LightMem (p<0.01 after Holm correction). It obtains the highest observed scores on MemoryAgentBench although the relative difference is low. Ablations identify controlled fragmentation and schema-aware assignment as the largest contributors to answer quality. Additionally, a 700-question cross-model evaluation with GLM-4.7 and Gemma-4-31B supported model independence. These results show that long-term memory can be better handled as constrained state management rather than continual summarization.

Large language models (LLMs) are increasingly relied upon to support ambient documentation and clinical reasoning. Here we examine the impact of a failure mode shared between these two applications by assessing their sensitivity to information incidental to the patient encounter. In 576 patient-clinician dialogues, we found that frontier models inserted small-talk exchanges into 35% of notes, while mean quality scores changed by at most 0.20 points on five-point scales. In 3.7% of frontier notes, models misattributed the asides or used them clinically. In 57 mock recorded consultations, background speech from a separate patient encounter at -10 dB leaked into 48.2% of transcripts, with contamination detected in 5.3% of downstream notes generated by four open-weight models. We propose a dual encoding hypothesis of clinical reasoning and distraction in LLMs, with preliminary evidence that LLM components associated with disruption by incidental information also support clinical reasoning. These findings support evaluating resistance to incidental information before clinical use, with safeguards that prevent contamination while preserving clinical reasoning.

We present fourteen TESS Objects of Interest (TOIs) from the SOPHIE-TESS follow-up program. Combining TESS photometry, SOPHIE radial velocities (RVs), ground-based photometry, and high-resolution imaging, we establish the nature of all targets. Nine systems (TOI-6457, TOI-6883A, TOI-6649, TOI-6842A, TOI-5217A, TOI-6893A, TOI-7025, TOI-7176, and TOI-5893) host planets. TOI-6649 b, TOI-6842A b, and TOI-5217A b are transiting hot Jupiters with periods of 4.4-6.0 d, radii of 1.3-1.6 RJ, masses of 0.3-1.3 MJ, and densities of 0.1-0.6 g/cm3. TOI-5217A also hosts an outer companion, TOI-5217A c, with a period of about 1761 d, a minimum mass of about 13 MJ, and an eccentric orbit (e = 0.53 +/- 0.02). TOI-5893 b, TOI-6893A b, TOI-6457 b, TOI-7176 b, and TOI-7025 b are transiting warm Jupiters with periods of 40.4-67.0 d, radii of 0.8-1.2 RJ, and masses of 0.5-1.0 MJ. TOI-6883A b, a recently reported warm Jupiter, is further characterized with new photometry and SOPHIE and TRES RVs. The other five systems (TOI-5621AB, TOI-5321AB, TOI-3678AB, TOI-5906AB, TOI-5308AB) show large RV amplitudes indicating eclipsing binaries (companions of 0.1-0.6 Msun and 0.1-0.6 Rsun). TOI-6842A b is a super-puff (density 0.08 +/- 0.01 g/cm3), among the nine lowest-density planets known, and with a transmission spectroscopy metric of about 250 it is an excellent target for atmospheric characterization. The six warm Jupiters, with equilibrium temperatures of 377-846 K, are low-irradiation benchmarks for hot-Jupiter inflation models. TOI-7025 b has an eccentricity of 0.82 +/- 0.02, one of the largest known for a planet orbiting a single star. Finally, using Gaia astrometric excess noise, we detect TOI-5906 B, the first object identified by transit, RV, and astrometry.

In this article, we study reduced even and odd Khovanov homology, with a particular focus on positive and $0$-adequate links. For positive links, we show that the first reduced homology is supported in a single quantum grading, where it is free abelian, and that its rank is determined by the Seifert graph of any positive diagram. As a consequence, reduced even and odd Khovanov homology, as well as unreduced odd Khovanov homology, detect fiberedness among positive links, extending the previously known result for unreduced even Khovanov homology. We further show that the torsion in these three theories does not detect fiberedness in the same way as the torsion in unreduced even Khovanov homology. We further show that $(p,q)$-cables of positive knots with $q\geq p$ exhibit the same behavior, even though such cables need not themselves be positive. We also study the stable homotopy types associated to reduced even and odd Khovanov homologies in the extremal and almost extremal quantum gradings of $0$-adequate links. At the almost extremal quantum grading, we show that the reduced stable homotopy types are wedges of sphere spectra. This contrasts with the unreduced setting, where the corresponding homotopy types depend on whether the $0$-state graph is bipartite.

We compute the temperature dependence of effective spin-model parameters in two-sublattice ferrimagnets with antiparallel and inequivalent sublattice magnetizations using Green's-function theory. The effective intra- and intersublattice exchange interactions, together with the sublattice anisotropies, exhibit distinct thermal renormalizations arising from the different magnetic interactions and correlations of the two sublattices. Although these effective interaction parameters decrease monotonically with increasing temperature, their unequal renormalization produces a strongly nonmonotonic and polarization-dependent long-wavelength spin-wave response. In particular, the spin-wave exchange stiffness exhibits a strongly branch-dependent and nonmonotonic temperature dependence. One polarization develops an enhanced spin-wave stiffness in the vicinity of the angular momentum compensation region despite the monotonic thermal reduction of the underlying exchange interactions. The Green's-function predictions are in quantitative agreement with atomistic spin-dynamics simulations, which provide a numerical benchmark including thermally induced spin correlations within the underlying atomistic spin model. These results establish a direct connection between temperature-dependent atomistic interactions, long-wavelength spin-wave stiffness, and the effective parameters required for finite-temperature micromagnetic and multiscale modeling of ferrimagnetic materials.

We construct non-co-sofic invariant random subgroups in connected simple real Lie groups of rank one with finite center, in Zariski connected simple rank-one groups over non-Archimedean local fields, and in Aut(T) for regular and bi-regular trees T. This gives a negative answer to the continuous analogue of the classical Aldous-Lyons conjecture for these groups. Our constructions build on the existence of a finitely generated non-sofic group and answer a question raised in the author's ICM paper [Gel18].

Superconducting circuits have emerged as one of the most promising and scalable platforms for quantum com- puting, with substantial industrial adoption driving qubit counts steadily upward. Yet the layout of the corresponding chips is still primarily performed manually, consuming multiple days or weeks of expert effort, even for moderately sized designs. Existing automation approaches sidestep this bottleneck by simplifying the underlying problem and relaxing physical constraints. Therefore, they fall short of fabrication-ready results. To overcome this scalability wall without such compromises, we first introduce a formal abstraction that translates physical device characteristics into a rigorous geometric problem formulation. This abstraction bridges physical realization and design automation. Building upon this abstraction, we then propose a three-stage design automation flow comprising geometry-aware global partitioning, ILP-based port assignment, and a hierarchical routing pipeline. Evaluations conducted by our interdisciplinary team of design au- tomation researchers and superconducting hardware experts con- firm that the resulting flow produces complete, manufacturing- ready layouts that comply with the physical design rules at the push of a button. That is, it operates fully automatically within seconds to a few minutes, and solves complex instances up to 25x faster than the state of the art. All methods are released as the fully open-source tool mqt-scpd as part of the Munich Quantum Toolkit, providing the first readily usable, end- to-end design flow for planar superconducting quantum chips that explicitly addresses the underlying physical characteristics and requirements.